Bug 16336 - last update of mageia4 render further updates impossible
Summary: last update of mageia4 render further updates impossible
Status: RESOLVED INVALID
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 4
Hardware: x86_64 Linux
Priority: Normal major
Target Milestone: ---
Assignee: Mageia Bug Squad
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-07-09 23:22 CEST by gilles d
Modified: 2015-07-11 19:45 CEST (History)
2 users (show)

See Also:
Source RPM:
CVE:
Status comment:


Attachments
exact list of installed mageai4 packages before attempting the dangerous update (258.87 KB, text/plain)
2015-07-09 23:22 CEST, gilles d
Details
result of journalctl -a --since="2015-07-08" > journalctl.txt + xz journalctl.txt (92.39 KB, application/octet-stream)
2015-07-11 12:08 CEST, gilles d
Details

Description gilles d 2015-07-09 23:22:59 CEST
Created attachment 6830 [details]
exact list of installed mageai4 packages before attempting the dangerous update

On Jul 8, 2015, the update applet signals that new updates are available.
The update contains ~100 meg of files, noticeably gcc 4.9.2 release 4.mga5 and updates rpmdrake, aria2, perl, python-rpm, etc.
At some point, installer hangs up and disappears.

Impossible after that to run mcc or any other urpm* utility. net_applet disappears. mcc etc segfault immediately. 


How reproducible:
I have a dual boot with identical copies of mageia4 (so no stress) and 
here is rpm -qa --last  of the state before the fatal update.
I suppose that by installing exactly all these packages and doing an upgrade should reproduce the problem.
Comment 1 Marja Van Waes 2015-07-10 22:58:00 CEST
Comment on attachment 6830 [details]
exact list of installed mageai4 packages before attempting the dangerous update

Can you please run (as root) the following commands in the install where updating went wrong

  journalctl -a since="20150708" > journalctl.txt

  xz journalctl.txt

and then pleas attach journalctl.txt.xz to this bug report

CC: (none) => marja11
Attachment 6830 mime type: application/octet-stream => text/plain

Marja Van Waes 2015-07-10 22:58:39 CEST

Keywords: (none) => NEEDINFO
CC: sysadmin-bugs => thierry.vignaud
Component: Release (media or process) => RPM Packages

Comment 2 Marja Van Waes 2015-07-10 23:00:30 CEST
Oops

It should be: --since="20150708"

please add the "--"
Comment 3 gilles d 2015-07-11 12:08:31 CEST
Created attachment 6834 [details]
result of journalctl -a --since="2015-07-08" > journalctl.txt + xz journalctl.txt
Comment 4 gilles d 2015-07-11 12:12:19 CEST
Attached above.
Crash is after removing old rpmdrake and restarting it.
Comment 5 Marja Van Waes 2015-07-11 12:53:19 CEST
I had not understood you were trying to upgrade.

I cannot see that you tried to update on the 8th of July, only that you tried to upgrade, and I see:

Jul 08 22:57:26 localhost.localdomain mgaapplet-upgrade-helper[6626]: adding media for new distribution failed


(In reply to gilles d from comment #4)
> Attached above.
> Crash is after removing old rpmdrake and restarting it.

but that's today:

Jul 11 11:59:40 localhost.localdomain drakrpm-update[12585]: [RPM] rpmdrake-6.20-1.mga5.noarch installed
Jul 11 11:59:40 localhost.localdomain drakrpm-update[12585]: [RPM] rpmdrake-6.10.5-2.mga4.noarch removed
Jul 11 11:59:41 localhost.localdomain drakrpm-update[12585]: restarting rpmdrake
Jul 11 11:59:41 localhost.localdomain mgaapplet[3884]: Computing new updates...
Jul 11 11:59:41 localhost.localdomain mgaapplet[3884]: running: mgaapplet-update-checker
Jul 11 11:59:41 localhost.localdomain mgaapplet-update-checker[17435]: ### Program is starting ###
Jul 11 11:59:41 localhost.localdomain mgaapplet[3884]: urpmi database locked, skipping updating urpmi database
<snip>
Jul 11 12:01:01 localhost.localdomain kernel: crond[19341]: segfault at 968 ip 00007f311ffbecbe sp 00007ffc11492e80 error 4 in libpthread-2.20.so[7f311ffb9000+17000]

(other than that, there's a bunch of "kernel: sd_festival" segfaults in the log)


Is this the same install, or the identical one?
Comment 6 gilles d 2015-07-11 19:05:24 CEST
Sorry not to having been clear.
It is an 'identical' one.
I have two identical bootable mageia4 partitions. absolutely identical except the UUID of the partitions they boot on.
One did fail on last upgrade, hence this report. 
What I did was to copy the second partition (state before upgrade) to the first, reestablishing a 'working' mageia on both partitions. This was a few hours after the problem, and before my report.
To answer your request, I just replayed the install, it failed all the same of course, and here is the journalctl. So the time tag is today and not 3 days ago.


Best
Gilles
Comment 7 gilles d 2015-07-11 19:19:51 CEST
Hi,

I've checked why mcc etc were crashing after update. I found that there was an old libreadline.so in /usr/local/lib that missed a symbol. (I remember having needed it at a time for some research code to work). The library was linked *before* the official one.

Removing this library *prevents the crash* during the install. Everything works fine.

For one time I make a report, I'm ashamed it is "my fault". To be true, there had been many many updates ok with this library in the line, this one was one too many.

Sorry again for bothering you, and thanks for Mageia, I love it!
Best
Gilles
Comment 8 Marja Van Waes 2015-07-11 19:45:39 CEST
Thanks for all the feedback, Gilles.

There is no need to be ashamed at all, we're all humans and we all make mistakes from time to time ;-)

I'm glad you found the cause and that you told us :-)

Keywords: NEEDINFO => (none)
Status: NEW => RESOLVED
Resolution: (none) => INVALID


Note You need to log in before you can comment on or make changes to this bug.