Text Based Installer Project » integration_tests
en

integration_tests

Text Installer Integration Tests

Active Bugs
Verified Bugs
Moved to Bugzilla

Verify for all installations:

  • Reboots to installed image
  • install log copied over
  • timezone set correctly
  • date/time set correctly
  • user and root created, root as role
  • nwam enabled/disabled
  • install on slice or partition, make sure everything else is preserved
  • correct language is enabled after install (check /etc/default/init after
    reboot)

SPARC

Test DescriptionImageTest MachineTesterStatusbuild 131 
my install testnightly1215mysparc_machinejoe testerbug 0 In progress
Install on whole disk text-install-2009-12-22-b.iso  Ultra 45  Keith  Success pass (Karen)
Install to existing specific slice, same sizetext-install-2010-01-06-04:32:21.iso  line1-ultra45 with 1.5 TB disk Mary  bug 45 pass (Karen)
text-install-2010-01-08-04:29:57.iso line1-ultra45 with 1.5 TB disk Mary  bug 53b  
text-install-2010-01-07-10:40:13.iso line2-t1 with 36 GB disk  Mary  Success 
text-install-2010-01-11-04:23:23.iso line1-ultra45 with 1.5 TB disk Mary    PASS  
Install to existing specific slice, change sizetext-install-2009-12-22-b.iso  Sun blade 1000 Karen  Successpass (Karen)
Verify all 4 root/user configurations    root/no user: Works (Keith)
root/user: works, root can't login(Karen) 		 * root/no user -pass(Karen)
* root/user-pass(Karen)
* no root passwd/create user - pass(Karen)
* no root passwd/no user - pass(Karen)
Install on nevada machine w/ufs, preserve ufs slice and install on a different slicetext-install-2010-01-08-04:29:57.iso  line1-sb1500  Mary    PASS - bug 22  
Install on s10 machine w/ufs, preserve ufs slice and install on a different slicetext-install-2010-01-07-10:40:13.iso Ultra 45  Sue  PASSED (had to workaround bug 31) INPROG (Keith) 
Install on nevada machine w/zfs, preserve zfs slice and install on a different slice     In progress (Karen)
Install on s10 machine w/zfs, preserve zfs slice and install on a different slice text-install-12-23.iso (nightly) SUNWA,70  Keith  Failed. Old rpool could not be imported under a different name - listed as corrupted.New rpool did not mount rpool/export/home on reboot. Possibly installed into the wrong slice (I think I selected s5, but s4 is where the installed system is at)
Install on nevada machine w/svm mirror      
Install on s10 machine w/svm mirror      
Install on machine with resizing of slice  text-install-2010-01-08-04:29:57.iso   line1-sb1500  Mary  PASS - bug 22  
Install on machine, leaving existing slice alone, install on new slice text-install-2009-12-22-b.iso  SunBlade 1000Karen  pass pass(Karen)
On machine > 2disk, install on other disk, ensure reboot to correct disk  text-install-2009-12-22-b.isoSunBlade 1000 Karen Picked slice 7 of 2nd disk.  It works.  pass(Karen)
Install on slice, make sure other slices are preservedtext-install-2009-12-22-b.iso  SunBlade 1000Karen   pass
Can install be interrupted     
Can install be restarted after being interrupted?      
Can install be restarted after successful, but not rebooted?       
Is language set correctly text-install-2010-01-07-10:40:13.iso laptop  Sue  PASSED  inprog (Sue)
Review online Help Text  Sue  Complete, sent comments to Barbara 12/11  
ldom testing text-install-2009-12-16-04:18:51.iso  line1-t5120-ldom  Mary  Failed - ICT fail to install boot loader  
text-install-2010-01-11-04:23:23.iso line1-t5120-ldom  Mary  Failed  bug 64   
text-install-2010-01-12-04:24:08.iso line1-t5120-ldom  Mary  Failed  bug 65   
Memory testing -512 MB text-install-2010-01-07-10:40:13.iso  line2-sb100 Mary  PASS - bug 22   
Memory testing -768 MB text-install-2010-01-07-10:40:13.iso  line2-v120 Mary  PASS - bug 22   
Memory testing -896 MB  Mary   
Memory testing -1024 MB text-install-2010-01-08-04:29:57.iso  line1-sb1500 Mary  PASS - bug 22   
Memory testing -1536 MB  Mary  
Memory testing -2048 MB text-install-2010-01-07-10:40:13.iso  line2-t1 Mary  PASS - bug 22   

X86

Test DescriptionImageTest MachineTesterStatusBuild 131
my install testnightly1215myx86_machinejoe testerbug 0 In Progress
Install on whole disk text-install-2009-12-16-03:07:02.usb Extra laptop   Jean  timezone/log file/network issues results  SUCCESS
 text-install-2009-12-18-03:07:50.usb  Extra laptop  Jean  timezone not working, stuck at 100% results   
 text-install-2009-12-19*.usb  Extra laptop  Jean  PASSED   
 text-install-2009-12-23-03:14:46.usb  Extra laptop  Jean  PASSED  
Install on whole disknightly12-16  vbox   Sue  Succeeded   
Install to entire existing sol partition text-install-2009-12-16-03:07:02.usb  Extra Laptop  Jean   timezone/log file/network issues results   SUCCESS 
 text-install-2009-12-18-03:07:50.usb  Extra laptop  Jean  timezone not working results  
 text-install-2009-12-19*.usb  Extra laptop  Jean  PASSED 
 text-install-2009-12-23-03:14:46.usb  Extra laptop  Jean  Failed. Failed to initialize disk. 
 text-install-2009-12-23-b.usb  Extra Laptop  Jean  PASSED 
 text-install-2009-12-23-b.iso  vbox  Alok  Failed bug 37  
 text-install-2010-01-06-03:39:11.iso  line1-ultra40 with 1.5 TB disk Mary  Failed bug 46  
Install to existing sol partition, change size text-install-2009-12-16-03:07:02.usb   Extra Laptop  Jean  timezone/log file/network issues  Doesn't work with no user results  SUCCESS
 text-install-2009-12-18-03:07:50.usb  Extra laptop  Jean  timezone not working results  
 text-install-2009-12-19*.usb  Extra laptop  Jean  PASSED 
 text-install-2009-12-23-b.usb  Extra laptop  Jean  PASSED 
Install to existing specific slice, same sizetext-install-2009-12-16-03:07:02.iso vbox Alok   Fail bug 13  
  text-install-2009-12-23-b.usb  Extra laptop  Jean  PASSED but this was on slice 0 SUCCESS 
  text-install-2010-01-07-12:13:29.iso  line1-ultra40  Mary  Fail on system with 1.5 TB disk bug 53a  
  text-install-2010-01-11-03:21:49.iso  line1-ultra40  Nary  PASSED  
Install to existing specific slice, change size text-install-2009-12-16-03:07:02.isovbox Alok   Fail bug 13  success (Karen)
  text-install-2009-12-23-b.usb  Extra laptop  Jean  PASSED but this was on slice 0 SUCCESS 
Install to different slice, change size  text-install-2009-12-23-b.usbExtra laptop  Jean  PASSED - installed on s3 SUCCESS
existing linux swap partition and sol1 partitiontext-install-2009-12-23-b.iso vbox Alok  Fail bug 35   
existing sol partition, remove and create new one in diff location 2010-01-07  extra laptop  Jean   PASSED SUCCESS
Install onto new sol logical partitiontext-install-2009-12-16-03:07:02.isovbox Alok   
  2010-01-07 extra laptop  Jean  PASSED SUCCESS
Install onto existng sol logical partition2010-01-07 extra laptop  Jean  PASSEDSUCCESS
Verify all 4 root/user configurations build 22-b.usb  Extra Laptop  Jean  PASSED SUCCESS 
Install on nevada machine w/ufs, preserve ufs slice and install on a different slicetext-install-2009-12-24-03:07:43.iso vbox Alok  PASSED  
Install on s10 machine w/ufs, preserve ufs slice and install on a different slicetext-install-2009-12-24-03:07:43.iso vbox Alok  S10 install stuck at 95%  
 text-install-2010-01-08-03:34:12.iso line1-x4200m2  Mary  PASS bug 57  
Install on nevada machine w/zfs, preserve zfs slice and install on a different slicetext-install-2009-12-24-03:07:43.iso vbox Alok  Text installer not able to see the free space outside the zfs pool SUCCESS (Keith)
Install on s10 machine w/zfs, preserve zfs slice and install on a different slicetext-install-2009-12-24-03:07:43.iso vbox Alok   
Install on nevada machine w/svm mirror       
Install on s10 machine w/svm mirrortext-install-2009-12-16-03:07:02.iso  line2-x4100  Mary  Fail - the install should fail but instead it install the whole disk - Bugid 6191  bug 15  bug 16  bug 17  
Install on machine with Linux et al on a different primary partition    Keith    Success* (*Ubunutu 9.10's grub is installed to the MBR and was clobbered by the install; setting up the system to be fully dual boot would involve some fiddling with Ubuntu's grub2 and OpenSolaris' grub. However, the Ubuntu partition itself was unharmed) 
Install on machine with Linux et al on the partition and it needs resizing    Keith   SUCCESS
Install on logical partition, Linux et al on a primary       SUCCESS  
On machine > 2disk, install on other disk, ensure reboot to correct disk       
Install on partition, make sure other partitions preserved   Keith    
Install on slice, make sure other slices (and partitions) are preserved       
For existing sol1 partition, make sure it's now solaris2 partition    Keith    FAIL if the size remains the same SUCCESS if you resize.
Can install be interrupted       
Can install be restarted after being interrupted?        
Can install be restarted after successful, but not rebooted? nightly12-16 vbox  Sue  Failed, ZFS part of TI failed (ti_install:195), send mail 
Is language set correctly text-install-2010-01-07-12:13:29.iso laptop  Sue  PASSED  SUCCESS
Review online Help Text  Sue  Complete, sent comments to Barbara 12/11 
vbox testing       
ilom testingtext-install-2009-12-16-03:07:02.iso line2-x4100  Mary   Pass - this is on line2-x4100 and the navigation keys are working    
xvm testingtext-install-2009-12-16-03:07:02.iso  osol-bldx-domU  Mary   Fail - none of the up/down/left/right navigation keys are working  bug 14    
Memory testing -512 MBtext-install-2009-12-18-03:07:50.iso  linea-i3 Mary  Fail - cannot navigate if I use xterm.  When I switch to use vt100, text-install does not come up bug 28  bug 29   
 text-install-2010-01-11-03:21:49.iso  linea-i3  Mary  PASS - use default TERM and it works fine  
Memory testing -512 MBtext-install-2009-12-18-03:07:50.iso  line1-dell450 Mary  Fail - cannot navigate if I use xterm.  When I switch to use vt100, text-install does not come up bug 28  bug 29   
Memory testing -768 MBtext-install-2010-01-07-12:13:29.iso line1-hpdc5750 Mary  Pass - use default TERM and it works fine  
Memory testing -896 MBtext-install-2010-01-14-03:42:26.iso line1-hpdx2200 Mary  Pass - use default TERM and it works fine   
Memory testing -1024 MBtext-install-2010-01-14-03:42:26.iso  tcdell-01Mary  Pass - use default TERM and it works fine 
Memory testing -1536 MBtext-install-2010-01-14-03:42:26.iso  line1-mpc375 Mary  Pass - use default TERM and it works fine 
Memory testing -2048 MBtext-install-2010-01-07-12:13:29.iso line1-ap2000 Mary  Pass - use default TERM and it works fine 

Test Machines

BRM:

  • x86:   
    • osai-brm-x-2 
    • osai-brm-x-3 
  • sparc:
    • osai-brm-s-3 
    • osai-brm-s-2 
  • assorted laptops available

MPK: 

http://installzone.eng/test/labinfo/lab_info.html

Active Bugs

DateStatusBug IDREBug DescriptionpriorityComment
01/05/2010 Fixed bug_40a Dave  2Tb disks need to be handled(x86) MEDIUM  In test. Hope to push tonight or tomorrow. 2Tb limit code for x86 has been pushed.
1/25/2010: Asked test group to help verify.
01/13/2010Fixed  bug_71Jean/Karen  generate iso.sort file  HIGH-  Fixed for x86, not important for b131 

Fixed and Verified bugs

DateStatusBug IDREBug DescriptionpriorityComment
12/16/2009VERIF bug_1 install log not copied over Karen has pushed fix
12/16/2009VERIF bug_2 timezone not set Karen has pushed fix
12/16/2009VERIF bug_3 nwam always enabled after rebootHIGH expect to be fixed end of Friday.  Karen pushed fixed on Thurs, 12/17
12/16/2009VERIF bug_4 failed to create BE's shared dataset (rpool/export/home/) trailing slash in name. (possibly when no user created)HIGH Cause known (Jean)
12/16/2009VERIF bug_5 ICT copy_sparc_bootlst fails (sparc)HIGH tentative fix by Alok. Fix pushed by Alok on 12/17
12/16/2009VERIF bug_6 ICT update_boot_archive fails (all?)HIGH Cause unknown Alok. Fix pushed by Alok on 12/17
12/16/2009VERIF bug_7 "Reboot" option is available when installation failsLOW Consult with Frank to see if this should be there or not. Assigned to Sue.12/17 Frank said "no". Fix pushed.
12/16/2009VERIF bug_8 Badly labeled disks crash the disk selection screenHIGH Assigned to Keith. Fix by COB tomorrow
12/16/2009CAN'T VERIF Need to Doc bug_9 zpools should be checked for existence before creatingHIGH Assigned to Karen. Code by COB 12/21 but ZFS bug may still cause issues.  ZFS problem: bug 13180 (http://defect.opensolaris.org/bz/show_bug.cgi?id=13180 ) Unable to verify, problem also exists for GUI.  Workaround is for the user to reboot and start the installation over.
12/16/2009VERIF bug_10 language is always C Alok pushed fix today.
12/16/2009Not a Bugbug_12 Sue when trying to install to p2 it still installs Solaris to p1LOWInvalid? User confusion. Will test again when EP support is there. Question for Frank (if user changes second one to Solaris, should it be first or second). Sue: 1/13 update: GUI behaves similarly. Able to make "first" partition unused and install into "second" partition. After install, Fdisk shows that install was to partition 1, from cyl 1.
12/16/2009VERIF bug_13 install to slice fails when using non-slice 0  (Jean hope to have fix by 12/20)HIGH Engine hardcodes s0 to zpool creation. Karen sending Jean info on change. 12/17 AI team confirmed that install to non-0 slice works
12/16/2009VERIF bug 16 Error in install_log: system_info.py: 68 Could not read /etc/sysconfig/language Alok pushed fix today.
12/16/2009VERIFbug 17KarenError in install_log: install_utils.py: 16 blocksLOWInstall succeeds. Gives user uneasy feeling. - Karen pushed fix for this 12/17. Unpushed because it broke sparc. 1/04: Karen to fix. target push of 1/13.
12/16/2009INVAL bug 18 Build will fail with EP merge.HIGH Dave consulting with William.
Keith has fix will push with EP merge. William has fixed, won't be a problem.
Dave is creating image and testing. ETA 12/22
12/17/2009INVAL bug_19 Regression of http://defect.opensolaris.org/bz/show_bug.cgi?id=38  Keith: I see this on the build 129 liveCD as well, so it's not ours. I've filed http://defect.opensolaris.org/bz/show_bug.cgi?id=13547
12/17/2009VERIF bug_20 Minimum Size display needs more precision:LOW.image_info reports image size of 600 ish MB, for the smaller image created today. The UI displays the minimum as 0GB (it should show perhaps 0.7GB). The exact minimum, when displayed, should be rounded *up* to the nearest 10th of a GB as well.
12/17/2009VERIF bug_21 Progress Bar stuck at 100% (saw on vbox, full disk install) "Performing Installation Completion Tasks" (workaround: hit CR)
12/18 Also happened earlier in install, on failed install
Does seem to eventually come back but takes a long time.		MEDIUMThis is a result of the UI code getting confused by the system time getting changed. On most screens, the code for capturing keystrokes is blocking;during installation, the getch() call is changed to be non-blocking (times out after 100ms of no keystroke). The timeout check appears to check the system time; thus, when the date time is set, if the clock is set backwards, (even by a few seconds, for example, if the clock gets reset from 15 sec past to 0), the "non-blocking" call takes that much longer to return, causing the install to appear to hang. Possible solution in the works, depending on whether or not the threading module suffers from the same confusion if the system time changes.
12/18/2009VERIF bug_23 If both user and root pw set, can still login as root  Cause Known (Karen)
12/18/2009VERIF bug_25 Normal user error (wrong value in hour field) appeared as ValueError traceback in install_log Verified by Sue 1/04 
12/18/2009VERIF bug_26  If user leaves root pw left blank, confirms to go to summary screen, then returns to users screen, root pw fields each filled with 13 asterisks.MEDIUMNeed to modify when we encrypt the password to account for the case of leaving and coming back to that screen
12/18/2009Filed bug 13875
VERIF 		bug_27 "Default language" on summary screen shows en_US.UTF-8. UI Spec shows "English". LOWinvestigate how GUI does it.
Should be filed as a P3 bug.
1/11/2010VERIF bug_28Marysetting TERM with xterm does not work on linea-i3 and line1-dell450, key does not navigate properly.  Both of these system are hooked up with kvm switch to a View Sonic terminalLOW Keith to check with Mary
Update from Mary on 1/7/10:
I bypass the KVM switch and hooked up a terminal and keyboard to the two x86 systems that had 512 MB.  
default sun-color - key navigation does not work
xterm - key navigation does not work
vt100 - text install will become blank
I even try ESCDELAY=1000 and it does not help.  I believe this might be an issue on x86 with low memory.
Update from Mary on 1/11/10:
With the latest x86 iso and if I use the default, the key navigation works fine.  
12/18/2009Not a Bugbug_29Marysetting TERM with vt100 and text-install does not come up at all.  This only happen on linea-i3 and line1-dell450.  Both of these system are hooked up with kvm switch to a View Sonic terminalMEDIUM Keith to check with Mary
 12/22/2009 VERIFbug_30 If I modify the size of the slice to be slightly larger than the available size displayed, it is allowed.  For example, if available space is 10G, I am allowed to type 10.1. It correctly stops me on any larger sizes, for example 10.2 or or any larger value.  Despite this, the size eventually got rounded back down to 10G in the summary screen, and also the installation does not corrupt data.LOWMight confuse the user, rounding issue. 1/11: In some cases, the reverse scenario can also occur - the UI may display a minimum of 4.5GB and *not* accept a value of 4.5GB. However, setting to 4.6GB allowed the installer to continue fine. Both cases should be resolved when this bug is fixed.
12/22/2009 VERIF bug_31KarenUI does not make sure user specified enough space for installation.  I selected to use whole partition, and specified 1G for that partition.  UI doesn't catch this, and it should HIGH1/04 Karen to fix. Targeted to push 1/6 or 1/7. Should be in 1/9 image. Keith: Installation failed - it appears the minimum size calculation did not consider the need for swap. The following lines came up in VBox when attempting this install:
Installation Size: 4637 MB
Available Size: 4710 MB
Memory: 768 MB. Swap Required: True
Space required with swap: 5149 MB
12/23/2009 VERIF  bug_32 When chosing whole partition, fails to initialize disk. What's actually happening is that s0 is getting blown away for some reason.  HIGH 
12/23/2009 VERIF bug_33  Sue Default user's path is set to "/usr/bin:"  MEDIUM  1/04 Karen to investigate See bug 13445. (http://defect.opensolaris.org/bz/show_bug.cgi?id=13445 )Keith: I think this was from creating a user named 'jack', which had other problems such as no home directory (due to the inclusion of a 'jack' user on the install media). Short term answer is to not allow jack as a user. Long term answer is to refactor SUNWslim-utils to only contain livecd code and have the jack user stuff in a different package. Targeted to push by 1/8. Disallow Jack as user pushed by Sue on 1/6. Verified 1/7 respin.  Bug 13750 filed to request for separating jack user files from other live media files.
12/23/2009  VERIF  bug_34KeithRunning in virtual box, this particular virtual box hard disk must be in a weird state from some previous testing/debugging that I was doing.  When the UI comes up, instead of showing me the list of partitions, it showed me the list of slices on the hard disk.  However, all the text on the screen talks about them as being partitions.  I clicked next, and it says that the partition table is invalid, and it wants to propose a partition table, and again, it lists all the slices...  HIGH   Very confusing.  I didn't go on further.  I preserved that virtual machine instance for Keith to debug.  I am using text-install-2009-12-23-b.iso - Karen
1/04/09 - Keith noticed something similar today. Karen to look at clean vbox disk using fdisk and format.
1/04/09 - Karen update 1/4/2010: tried to run format command, I can't run it.  It reports "Failed to initialise libfdisk".  Then, I ran test_td program, and got the following 2 output:
file:///net/ox.sfbay/code/discover_disk_part.png
file:///net/ox.sfbay/code/discover_slice.png
Then, I ran "fdisk /dev/rdsk/c2d0p0" which asked whether I want to create a partition that's 100% for Solaris.  I selected yes.  After that, I am able to use the format command, and saw that there's one active Solaris partition.  Then, I use the test_td program again, the -p option also reports one partition.  Ran the text installer, and now, everything looks right, and the install is able to complete successfully.  So, looks like not having any partition on the disk confused the code somehow. Targeted push by end of week (1/8)
12/23/2009  VERIF bug_35  text-install-2009-12-23-b.iso
Partition 0 is Linux swap.
Partition 1 is Solaris 1.
I selected partition 1 to have the 'rpool' and it blew up
with this stack trace:
2009-12-23 05:44:58,404 - ERROR   : text-install:258 Traceback (most recent call
  last):
      File "/usr/bin/text-install", line 228, in <module>
          screen = screen.show(install_profile)
        File
      "/usr/lib/python2.6/vendor-packages/osol_install/text_install/base_screen.py",
      line 103, in show
          return self.validate_loop()
        File
      "/usr/lib/python2.6/vendor-packages/osol_install/text_install/base_screen.py",
      line 131, in validate_loop
          self.on_continue()
        File
      "/usr/lib/python2.6/vendor-packages/osol_install/text_install/partition_edit_screen.py",
      line 200, in on_continue
          solaris_part = self.install_profile.disk.get_solaris_data()
        File
      "/usr/lib/python2.6/vendor-packages/osol_install/profile/disk_info.py",
      line 247, in get_solaris_data
          raise AttributeError, "No Solaris data on this disk"
The other thing to note here is that the UI did not report
the sizes of partition 0/1 at all. Had 'em set to '0'. 		HIGH Alok reported this.
1/04 Alok to reproduce to see what Linux swap and Solaris partition show up as on partition screen. Dave to email William about determining partition.
1/5 Alok reports that both the Linux swap as well as Solaris1 partition show up as Solaris2 partitions. Keith to backout changes that convert solaris1 to solaris2 partitions early on. Once done, need Frank to see if UI ok.
12/23/2009INVAL bug_36 I built an image with limited package list.  Most likely doesn't include the language packages.  I booted up the image, the select desktop  language question come up, and there is no language listed on the list.  It says that the default is 0.  I hit return, got back the select desktop language prompt again.  I typed in some random number, also got the prompt again.  Looks like there's really no way out of this unless I build an image with some language packages.LOW  - reported by Karen. Result of custom images. Update by Alok: this is certainly a pre-existing bug. Defect 519
12/23/2009 VERIF  bug_37JeanCreate a Solaris2 partition as the first partition on the disk and have it occupy 90% of the disk. Set hostname to 'xx', 'None' networking, edit 'Time' manually, set root password to 'x' and leave out user details. Watch the Install blowup with TypeError: ZFS part of TI failed. No s0 created. HIGH  - reported by Alok. 1/04 Karen to gather data: 1. How does gui installer handle if no partitions are active? 2. How does gui handle if there's another active partition?
In GUI, all partitions on the disk are left as is.  GUI does not explicitly make any partition active at target instantiation time.  Even partitions that are newly created/modified by libti does not have the active flag set.  The partition used for OpensSolaris is made the active partition by the ict_installboot ICT, which calls the installgrub command to set the active flag.  We should do the same.  Changes needed in the PartitionInfo.to_tgt() function to not set the active flag for modified partitions.  Changes also needed in tgt.so code to not require the active flag.  Right now, it fails if the active flag is not set for a partition.
01/05/2010VERIF bug_38 Need to add EP supportHIGH1/04 Seems like we're not sending the right thing to format/fdisk, failing there. Trying to figure out why.
01/05/2010VERIF bug_39 Jean x86 shouldn't use happy face / graphical boot MEDIUM  Since we're installing in a text environment, we can't assume that graphical/happy face boot grub features are supported or desired. The installed machine should have a boot process that is 'boring'/black&white like the media itself does. Modify the install-finish script to not enable happy face boot for text install. Estimated push, 1/12
01/05/2010 VERIF bug_41 Correctly display exported zpools HIGH  Display pool name instead of "exported_zpool" for Zpools in the exported state
01/06/2010 VERIF bug_45 using existing slice to install fails on sparc  HIGH  Karen is looking at this on line1-ultra45 with 1.5 TB disk, whole disk install works on the same system.  Karen's update: The fix integrated on Jan 5 for fixing bug 34 has a bug, and caused any SPARC disk with existing slices to not get recognized correctly.  Keith will back out the fix for bug 34 since it does not really fix that problem either.  Any SPARC image prior to 1/5 should work in terms of installing into slices. Update from Mary on 1/6/10:
On line1-ultra45 with the 1.5 TB disk, I try to install text install images based on 12/24, 12/26 and 1/5 and they all fail when I try to install an existing slice.
I will try to install on sparc system that does not have 1.5 TB disk and update status tomorrow.
Update from Mary on 1/7/10:
With the latest sparc image from 1/7/10, I still cannot install into an existing slice on line1-ultra45 with the 1.5 TB disk. However, I can install into an existing slice on line2-t1 which had a 36 GB disk.  So the problem only occcurs on 1.5 TB disks.
01/06/2010VERIF bug_46 Alok  Need to ensure only one rpool on the disk  HIGH  Related to active zpool vs exported zpool. 1/7: Now that Jean has fixed bug_41, on my laptop, Select Slice screen comes up with slice0 as rpool. Can't F2 unless I F5 to unused and F5 again back to rpool. Keith says the fix for this bug will resolve that. Anticipated push on 1/8.
01/07/2010 VERIF bug_47 Keith Can't cycle through partition type if EP is empty  HIGH  during an install set up the system so that p1=extended with no logicals, p2 = Solaris 2. Install. Then start the text installer again. On the select partition screen, note that no logicals are listed and try to change the extended partition to Solaris 2. The text installer goes back to the menu screen (or cores).
01/07/2010 NOT A BUG  bug_48   rpool still exists  LOW-MEDIUM during an install set up the system so that p1=extended with no logicals, p2 = Solaris 2. Install. Then start the text installer again. On the select partition screen, just take the defaults to install to the Solaris 2 partition. When the install starts, it dies in ti_install.py:192 Root pool rpool exists.This didn't happen when I rebooted between installs. this is expected behavior. User error.
01/07/2010 VERIF  bug_49   Better error messaging needed  MEDIUM  If I set up my machine via fdisk so that p1=FAT16 p2= Solaris 1 (active). I then fire up text-install and try to install on p2. The error message tells me I need a Solaris partition. In my mind there is one. A better error message is needed. If I create p3= Solaris 2 partition and try to install on that it fails. Dies when trying to fdisk the disk. Discussion with Alok brought up the idea that we might not want to boil the ocean here and should consult with Frank and possibly just beef up the error message.
Logs are at: /net/indiana-build.central/export/home/jeanm/bug_49
01/07/2010 Not our Bug bug_50 Karen  User path and home not set properly if none for networking  MEDIUM  If I select none for networking, and specify a user then when I try to login after the install I get path set to /usr/sbin and the message No directory! Logging in with home=/.  It is caused by a recent change in slim_source to fix bug 364.  User's home directory are now controlled by auto_mount maps, and user's home directory is specified as localhost:/export/home/&.  If nwam is not enabled, "localhost" is not recognized.  I can't not even ping 127.0.0.1.  So, mounting of the home directory for that user failed.  Even though nwam is disabled network/physical:default is enabled.  At this time, I don't know what other service need to be turned on for this to work.
This appears to be a Solaris bug and not text-install specific. (Evan, Sanjay, Dave(s), Clay have been diagnosing)
Solaris bug 6913780
01/07/2010 VERIF  bug_51  Alok Non-existent zpool found  MEDIUM  If I install nevada on a slice and into a pool named 'otherpool' and subsequently start the text installer, the installer not only discovers 'otherpool' but also discovers a non-existent pool named 'oldpool' of size '0'. 'zpool import' output shows a single pool named 'otherpool'. Alok has the vbox vmdk that reproduces this problem if anyone would like to take a look. 
01/07/2010 VERIF bug_52Jean Installing onto a partition that was an EP/LP with Solaris on it fails  MEDIUM  I first install onto a LP on an EP on p1. Works. Reboot. Then run the installer again. I change the LP to unused and the EP to Solaris 2. The install fails in the fdisk call.
01/11/2010 VERIF bug_53a Dave  Install on existing slice fails on x86 systems with 1.5 TB disk,  This works fine on systems that does not have 1.5 TB disks.  HIGH Verified as working with image on 1/11/10
01/11/2010 VERIF bug_53b  Karen  Install on existing slice fails on sparc  systems with 1.5 TB disk,  This works fine on systems that does not have 1.5 TB disks.  HIGH  Verified as working with image 1/11/10
# more install_log
2010-01-08 12:59:09,827 - INFO    : text-install:118  START
2010-01-08 13:09:31,503 - INFO    : network_type.py:181 Configuring NIC as: auto
matic
2010-01-08 13:10:00,049 - ERROR   : ti_install.py:271 Failed to initialize disk
2010-01-08 13:10:00,050 - ERROR   : ti_install.py:272 VTOC part of TI failed
Traceback (most recent call last):
  File "/usr/lib/python2.6/vendor-packages/osol_install/text_install/ti_install.
py", line 230, in do_ti
    tgt.create_disk_target(tgt_disk, False)
TypeError: VTOC part of TI failed
2010-01-08 13:11:46,629 - INFO    : text-install:99  END
#
Update on 1/11/10:
The install works fine with the image on 1/11/10.
01/11/2010 VERIF  bug_54 Dave  text install does not report the correct sizes for 1.5 TB disk on x86 only. This works fine on sparc  HIGH  Verified as working with image on 1/11/10
01/07/2010 DUP (bug_30)  bug_55  In the UI, where it checks for minimum required space.  If you specified exactly the amount of minimum space it displays, it will error and not allow it.  But if you specify min_space + 0.1, then, it will allow you to continue Probably another rounding issue??
01/08/2010 VERIF  bug_56 Keith/Jean  partition failure, fdisk fails, used to work? HIGH  Reported by Sue. Using vbox.
On partition screen, started with:
solaris2  16   0
unused     0   0
Changed solaris2 to unused. Showed 0 size, 16 avail.
Arrowed down to next partition entry. F5 gave me:
 Unused    0   1.0           Logical   size   avail
 Extended   15  16            Unused      0     15
Another F5 gave me:
 Unused    0   1.0           Logical   size   avail
 solaris2  15  16                                    15
(Shouldn't the 15 be cleared under Logical avail?)
Continued on and started install. Failed right away with:
install_progress.py:140 fdisk part of TI failed
Investigating to see if this is a dup of bug 47
This is a size issue. If I reduce the size of the Solaris 2 partition it works.
If I run fdisk by hand with the proposed partition table I get a message basically telling me it's too big.
I think the issue is with the Rsect number. It's set to 658665.
Keith: It looks like in some cases, the installer tries to create partitions whose size + offset goes past the end of the disk. PartitionInfo.adjust_offset() and PartitionInfo.to_tgt() both need to be examined to determine how to prevent this from occurring.
Sue: Fix in 1-11 image did not work.
Sue: Fix in 1-13 image caused traceback: AttributeError: 'tgt.Slice' object has no attribute 'id'
Sue has verified. Waiting for Frank to verify his scenario. Frank verified at 11:23am.
01/10/2010 VERIF  bug_58  Jean  System with 6 LP's cores in tgt discovery  MEDIUM  If I use create a system such that there are 6 LPs then the call to discover_target_data cores in TgtPartition_enumerate, TgtPartition_Create. fdisk partition file and core are in /net/indiana-build.central/export/home/jeanm/bug_58
01/11/2010VERIF bug_60 swap showing up as "mount" on slice screen  MEDIUM+ found on Sue's laptop, running from latest 1/11 ws. dump_tgt shows Type: mount and User: swap. Sue: 1/12 verified from latest workspace
01/11/2010 VERIF  bug_62 Alok/Keith  Allow installation into something other than rpool  HIGH  Hope to test and maybe push tomorrow. Fix pushed 1/14
01/11/2010 INVAL  bug 64   text install in sun4v did not install successfully   (Keith: removed traceback. It's in the wiki history if needed)
Update on 1/12/10:
This bug is another manifest of bugzilla bug 13267 due to the fact that the ldom had multiple virtual isos.  Once I remove the ldom and reconfigure the ldom to use only one virtual iso, the install is successful and got pass the failure.
01/12/2010VERIF bug_66Jack Remove DDU entry from small menuHIGH When DDU is ready and tested, will be added back. Fixed 1/13. Sue: 1/14 verified
01/13/2010VERIF bug_68 Keith  Whole Disk installs fail starting with 1-13-respin images  HIGH  Fix has been pushed and will be available in 1/14 image. Sue: 1/14 verified
01/13/2010VERIF  bug_69 Keith  Allow CTRL-C to kill text installer on first screen only  LOW  In case of terminal issues, hitting F9 to quit and navigating from CANCEL to QUIT can be difficult or impossible. Allowing ctrl-c to kill the installer, but only on the first screen, allows the user a way out if needed. Sue: 1/14 verified
01/13/2010Not our bug  bug_70 Jean  manifest needs to reflect package name changes  HIGH  bug 13741
After talking with Sanjay, this will be fixed at a later date along with the rest of the DC manifests
01/13/2010VERIF  bug_72Keith Negative long can't be cast to unsigned  HIGH 1/13 reported by Sue, vbox, using 1-13 respin. Same steps as bug_56, but on slice screen. Fixed on 1-13 Sue: 1/14 verified
01/14/2010 VERIF  bug_73Keith Repeated cycling of partition type caused crash  HIGH 1/14: Start with:
Unused      0    6
Extended   10   16
Unused      0    0
Unused      0    0
Logical has
Solaris2    10   10
Change Extended to Unused.
Moved down a slot and F5 cycled there. After a couple of cycles, it crashed and I got:
An unhandled exception occurred.
    error: addch() returned ERR
Logs send via email.
1/19: Sue verified w/1/19 image		01/14/2010 VERIF bug_76Sue Incorporate Barbara's latest help files HIGH pushed on 1/15 
01/14/2010 VERIF bug_78Keith Code to scan for existing zpools should be done in the target discovery thread  LOW/MEDIUM  The calls to "zpool status" and "zpool import" take a little bit of time and introduce noticeable lag when moving from the disk select screen onward. Running these calls from the target discovery thread, and updating the code to cache the returned zpool names, would be better.
01/19/2010 Can't reproduce bug_82 JeanFailed to transfer bits to the target MEDIUM for now  From Angela, 1/15, partition_07. 1/19: Jean able to reproduce. May be related to bug_81
Angela was unable to reproduce this with the 01/19 image.
01/19/2010 Verif bug_80 Keith  >16 Logical partitions crashes UI  MEDIUM  Constant value was wrong. Fix in progress. Expected push by end of day 1/19
01/20/2010 VERIF bug_83 Keith  Installer added .textinstall to installed system  HIGH  Prevents beadm from working
01/21/2010 VERIF  bug_84  Keith  Installer crashes in call to DiskWindow.print_data()  HIGH  Root cause: *Info.get_description() does not always return a string object, which causes the call to curses.addnstr() to fail
01/21/2010 VERIF  bug_85  Keith  Logging statement raises uncaught exception when number pad is used for some terminals  HIGH 
01/21/2010 VERIF  bug_86  Dave  Not looking for slices if partition is not active  HIGH  verified with 1/22 image.
01/25/2010INVAL  bug_90  Sparc: first install Nevada with ZFS into a slice, then, install OpenSolaris into another slice, not sure whether it works   Worked ok afterall, once Karen counted correctly.

Bugs moved to Bugzilla or otherwise dealt with (status field documents what was done to it)
Bugzilla search

DateStatusBug IDREBug DescriptionpriorityComment
12/16/2009 Filed 13871bug_11Keith reboot after install causes small menu screen to come up again. Selecting reboot on that causes small menu screen again.MEDIUM Keith investigating, confusing for user. To be filed as a P3
12/16/2009Filed 13872, fix-b4-pbbug_14Keithnavigation keys are not working in xvm pv installMEDIUMWorks in ILOM but not DomU. Keith investigating.
 - When ssh'ed in from an OSol b129 machine, xvm install proceeded as expected and the final (whole disk) install completed as expected given current bugs.
 - Mary to try on Sol 10
 - For snv_126, this only fails when you use rsh.  If I use ssh, it works.
 - For s10u8 fcs, this fails with both rsh and ssh.  I did this from ins3525-svr which is running s10u8 fcs.
 - Not considered a stopper.
Update on 1/6/09 by Mary:
 I am able to get this to work with both ssh/rsh for snv_126 and s10u8 fcs as follow:
   export TERM=xterm
  export ESCDELAY=1000
The key does navigate properly now after doing the above.  The text install works but fails to reboot successfully due to DomU not able to restart. However, I can do xvm PV domU install with osol_130 AI without any problem and DomU can restart after reboot.  When I did this with the x86 text install images based on osol_129 on 12/17, I did not have any problem with Domu unable to restart and reboot.
- Possible solution: As part of terminal selection menu, if user changes terminal from sun-color to anything else, assume they are connected remotely and bump up the value for ESCDELAY to an appropriate value (500-1000 [needs to be tested]).
- This will be associated with the terminal issue bug. Should be filed as a P2
12/16/2009documented in limitationbug_15JeanTrying to do text install with SVM should fail, but it is sucessful if I use the whole disk install.  Look at bugzilla 6191 and the install should failMEDIUMWhole disk was the only install tried. Mary is seeing what AI and liveCD do.
liveCD - GUI installer core dumps
need to run test_td to see if we can see the svm inuse data. Add pop up if they try to use that disk.
Jean will look at this.
Should be listed on existing limitations page.
Existing bugzilla bug 6191 should be updated.
12/17/2009Filed 13872, fix-b4-pb bug 22Sue display is not correct on sparc tipline install if you use default sun-color.  Workaround is to use vt100 and it displays properly. Probably another case where we need to have the user pick the terminal they are running from ???. Maybe add menu entry to select terminal type. Sue to ask Frank about menu entry.
Should be associated with the terminal issue bug. Frank says the menu should be before the small menu. i.e. similar to keyboard and lang selection.
12/18/2009Filed 13874, fix-b4-pbbug_24 On slice screen, made slice 0 (which was entire disk) unused. Made slice 1 rpool with same size. Other slices then showed Avail size of -0.0LOW Rounding issue
Should be filed as a P3 bug as a more general rounding bug. (bug 30) Should be fixed before push to slim_source
Jean has a fix for the -0.0 bug on thunderfury in /export/home/jeanm/slim_0.0/usr/src/cmd/t*/o*/p*/disk_info.py  Will push it after images are made.
01/15/2010filed bug 14070 bug_40b Dave  2Tb disks need to be handled(sparc) MEDIUM  Awaiting sparc info from ???
A P2 bug should be filed
01/05/2010Filed bug 13876, fix-b4-pbbug_42 Partition/Slice Edit fields should stay right-justified during editing MEDIUM  Should be filed as a P3 with Keith as RE. Should be fixed before push to slim_source
01/05/2010Filed bug 13901bug_43 Manual network screen: Domain name needs basic syntactic validation VERY LOW  Manual Network screen not yet supported.
Should be filed as a RFE.
01/05/2010Filed bug 13902bug_44 Hostname doesn't support true max length of 256 VERY LOW  Hostnames longer than approx. 60 characters can't be used, since the EditFields don't yet support scrolling to the left/right
Should be filed as a P4
01/08/2010Filed bug 13903 bug_57 I install s10u8 UFS on the system with default install.  There is a ufs swap partition named "swap".  In the text install, the "swap" is shown as rpool.  This should be "swap" instead MEDIUM  When 62 gets fixed this has a workaround.
Should be filed as a P3.
01/11/2010Filed bug 13904, fix-b4-pb bug_59  Logical Partition display issues on partition edit screen  LOW  When manipulating the logical/extended partitions, the following display bugs can occur: (1) The "Avail" space column for the last logical partition is not always cleared when the rest of that line is deleted (for example, from deleting the extended partition, or from 2 unused logical partitions collapsing into 1); (2) Logical partitions not associated with an existing extended partition get marked destroyed when a newly created extended partition is cycled to something else. This only occurs if there was already an extended partition.
Should be filed as a P3. Should be fixed before we push to slim_source.
01/11/2010Filed bug 13867 bug_63   libtd modification to assume partition 130 (0x82) is Linux swap until we can prove otherwise   Keith writing bug against libtd. Release notes should recommend (but not require) that people should run fdisk to convert the partitions from Solaris 1 to Solaris 2. See bug 13867 
01/12/2010filed bug 14001 bug_65   cannot reboot ldom after text-install is sucessful MEDIUM   After install into a whole disk or existing slice, the ldom cannot boot from disk after successful install:
SPARC Enterprise T5120, No Keyboard
Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
OpenBoot 4.30.2, 1024 MB memory available, Serial #83488611.
Ethernet address 0:14:4f:f9:ef:63, Host ID: 84f9ef63.
Boot device: /virtual-devices@100/channel-devices@200/disk@0:a  File and args:
The file just loaded does not appear to be executable.
{0} ok devalias
viso                     /virtual-devices@100/channel-devices@200/disk@2
cdrom                    /virtual-devices@100/channel-devices@200/disk@1
vdisk1                   /virtual-devices@100/channel-devices@200/disk@0
vnet1                    /virtual-devices@100/channel-devices@200/network@0
net                      /virtual-devices@100/channel-devices@200/network@0
disk                     /virtual-devices@100/channel-devices@200/disk@0
virtual-console          /virtual-devices/console@1
name                     aliases
{0} ok
Keith investigating.
Should be release noted/bug filed if Keith's investigation points to that it's nothing obvious.
01/13/2010added info to bug 13874bug_67  Error message for available space conflicts with slice screen  LOW  When I verified bug 30 I noticed that it did give me the error message that my new size was too big. However 3 issues, 1) the new size was in whole G no decimal. i.e. I put in 186.1 and it told me my size of 186 G was too big. 2) It says that "The new size (186) is greater than the availalbe space (185)" But the slice screen states that 186.0 is available.
3)When I have available as 186.3 I can set the size to 186.4 and it doesn't complain.
Should be associated with the rounding bug.
01/14/2010filed bug 14002bug_75Keith  Can't cycle Extended Partition with logical solaris2 to solaris2  MEDIUM-LOW Keith says this should work. Workaround: need to change logical solaris2 to unused first.
 Should be filed as a P3
01/14/2010 Filed bug 13924 bug_77  Invalid existing disk configurations result in indeterminate UI behavor  MEDIUM  Both tgt and the UI make the unfortunate assumption that an existing disk's partition/slice layout is valid, which leads to indeterminate behavior. Example 1: If a disk has multiple Solaris2 partitions, the UI will allow the user to continue, but always select the first Solaris2 partition as the target, even if the user may expect otherwise. Example 2: If any slices other than the backup slice have any overlap, the UI selects one or more of the overlapping slices and sets it size to 0, destroying it and any data on it.
01/11/2010 Bugster 6918569  bug_61 Can't F5 cycle to rpool on slice screen  MEDIUM on Ultra45 in Keith's office, changed rpool to unused. Moved to next slice down and was unable to change that slice to rpool. When 62 gets fixed, this one has a workaround so moving to medium. Sarah investigating libdiskmgt. Moved to bugster, the root cause is a zfs issue where zpool_in_use returns the same data for two slices that start on the same cylinder (including s2 and whichever starts on cylinder 0)
01/21/2010 Filed bug 14047 fix-b4-pb   Installer doesn't preserve BOOT and ALT slices  HIGH  See 14047 for more details
01/14/2010filed bug 14071bug_74Alok Pkg server list for sparc needs verification  MEDIUM   A P3 bug should be filed.
01/19/2010filed bug 14072, fix-b4-pbbug_79Keith  Solaris2 partition should be marked destroyed if it has no slices  MEDIUM  A Solaris2 partition without slices defined will have any data on it destroyed when the new VTOC is written. Partitions in such a state should be marked destroyed on the partition edit screen.
A bug should be filed for this.
01/19/2010filed bug 14074bug_81 Jean Deleted partitions not always deleted  MEDIUM-LOW  From Angela, 1/15, partition_06.Configuration: Existing partitions, including Solaris2 partition. All partitions except the Solaris2 partition are destroyed in UI. After installation, fdisk table still has the 'destroyed' partitions. 1/19: May be related to bug_82
The problem is that since the Solaris partition has not been modified and we don't consider deleting a partition a modification, then we send libti info to preserve all for all partitions and it does nothing.
Not needed for putback
01/21/2010Filed 14083, 14084, 14085bug_87   Blank disk, install to logical solaris2 causes ValueError  HIGH  Reported by Dave:
  boot up go to shell
  format then choose fdisk to destroy all partitions
  quit after saving
  text-install
  cycle partition to be Extended
  create solaris2 logical
  carry on with install - and you will get ValueError: tgt.Slice() "type" must be one of the appropriate class constants, got "backup"
 Installing to a Solaris2 primary partition on blank disk works.
 Sue reproduced, sent install_log to Keith
Should be release noted under limitations
A P2 bug should be filed.
01/22/2010 Filed 14144  bug_88    memory fault on sun4u, reported by test  HIGH  Core should be available as of 1/22. We don't handle it when disk doesn't report head or sectors. P2 bug should be filed, release notes should point to bug
01/25/2010  bug_89   If the part isn't resized the change from Sol1 to Sol 2 doesn't happen  MEDIUM Keith thinks the problem is in the modify code.  Only have problem with having Solaris1 and Linux swap at the same time.  Need P?? bug which depends on 13867.  Keith will file bug and also update release notes. Related issues are tracked under bug 13924 and bug 14146
Tags:
Created by Susan Sohn on 2009/12/11 17:07
Last modified by Susan Sohn on 2010/01/26 22:38

Collectives

Project caiman Pages

XWiki Enterprise 2.7.1.34853 - Documentation