SIGSEGV in Kopano-core on db-server disconnect

Hello devs,

below SEGV appeared on nearly current core-master ea1984086f3 after restarting the maria-db server while kopano-server was up. I was not able to clearly reproduce the case, sofar, but there seems to be a relation.
Generally I’m not sure, wheater you’re interested in suchkind of reports inbetween release-cycles of community-builds.

Feel free to comment.
I’m building on/for CentOS6 with PHP7 support.
++umgfoin.

Wed Nov  8 11:58:53 2017: [ notice] Starting kopano-server version 8.4.90.804 (pid 9814)
Wed Nov  8 12:22:54 2017: [warning] SQL [00007661] info: Try to reconnect
Wed Nov  8 12:22:54 2017: [error  ] KDatabase::Connect(): database access error Unknown error code (0x80000007), mysql error: Can't connect to local MySQL server through socket '/var/lib/mysql/mysql.sock' (2)
Wed Nov  8 12:22:54 2017: [crit   ] ----------------------------------------------------------------------
Wed Nov  8 12:22:54 2017: [crit   ] Fatal error detected. Please report all following information.
Wed Nov  8 12:22:54 2017: [crit   ] Application kopano-server version: 8.4.90.804
Wed Nov  8 12:22:54 2017: [crit   ] OS: Linux, release: 2.6.32-042stab120.20, version: #1 SMP Fri Mar 10 16:52:50 MSK 2017, hardware: x86_64
Wed Nov  8 12:22:54 2017: [crit   ] Thread name: z-s:
Wed Nov  8 12:22:54 2017: [crit   ] Peak RSS: 47040
Wed Nov  8 12:22:54 2017: [crit   ] Pid 9817 caught SIGSEGV (11), traceback:
Wed Nov  8 12:22:54 2017: [crit   ] Backtrace:
Wed Nov  8 12:33:52 2017: [ notice] Starting kopano-server version 8.4.90.804 (pid 13293)

Hi @umgfoin ,

thanks for the report. Did you already bisect it down to this specific commit or could the culprit also be an earlier commit (since I don’t see a direct relation between this change and database interaction).

Hi @fbartels ,
it appeared somewhere between commit ea1984086f3 and 2c4961156a3.
Didn’t happen again but I also didn’t restart the db-server, since.

In the past, restarting maria-db never crashed kopano-server and I had a lot of restarts in the last weeks while investigating the high-db-load issue (which got solved with your recent query-optimizations).

++umgfoin.

Hi @umgfoin ,

I agree with you that just restarting (or even shutting down for a brief period) should not generate segfaults in kopano-server. I just tried to reproduce this on the latest nightly (8.4.90.908) but was unfortunately not able to do so.

Hi @umgfoin

Do you have coredump enabled? In that case it should be easy to get a real stack trace (would guess there is more information).

Bo

@bosim said in SIGSEGV in Kopano-core on db-server disconnect:

Do you have coredump enabled?

Hi Bo,
unfortunately not.
'changed that and will return with stacktrace, if it comes up again.
++umgfoin.