View Issue Details

IDProjectCategoryView StatusLast Update
0003155unrealircdpublic2006-12-28 09:42
ReporterBricker Assigned To 
PrioritynormalSeveritycrashReproducibilityunable to reproduce
Status closedResolutionopen 
Product Version3.2.6 
Summary0003155: Random crash after delink (reroute, reroute, sync)
DescriptionSo heres the situation:

Our hub had a routing issue and was going in loops, so we relinked all the servers to a backup hub. Once the hub became available again (20-30minutes) we squitted the servers and rerouted them back to the hub. After that we had two, maybe 3 random IRCd crashes. Nothing in the error logs except a sync issue of smth like a few seconds off. We couldnt figure out what caused this and the best I can come up with is something having to do like

<IM LINED HERE AND SYNC'd> <NOW IM LINKED HERE AND SYNC'd> <NOW IM BACK HERE AND SYNC'D> <Wait, I'm not truly synced from the original! DIE> <--basically


Server1 LINKED/SYNC'd HUB
Server1 DELINKED HUB LINKED/SYNC'd BACKUP
Server1 DELINKED BACKUP LINKED/SYNC'd HUB
something of that sort, not sure if it makes any sense to anyone else...but just wondering if anyones ever heard of such a crash like that?
3rd party modules

Activities

stskeeps

2006-12-21 02:48

reporter   ~0012869

.. no coredump?

syzop

2006-12-21 07:43

administrator   ~0012871

Last edited: 2006-12-21 07:44

Indeed. Without a coredump / backtrace, there's little we can do.
When you report a bug, you should have seen instructions on what to do in case of a crash bug. Anyway....

Basically, try this:
./unreal backtrace
Then choose a nice core file (if there's something to choose, that is), and copypaste the output of above command back to us.

Also, let us know if you run any 3rd party modules (anything besides commands.so and cloak.so).

PS: This will not work if you recompiled or reinstalled the ircd after the crash

Bricker

2006-12-21 10:29

reporter   ~0012876

[gmbricker@zanj Unreal3.2]$ ./unreal backtrace
Core files available:
ls: *core*: No such file or directory
No core files found... Nothing to do

If you are sure UnrealIRCd crashed, then verify that unreal
has permission to dump core (type "ulimit -c unlimited" and see
if you get permission denied errors). Also verify that you did
not run out of quota.
If all that is ok, then it might be that Unreal did not crash but
got killed by the OS (eg: cpu/mem resource limits), the syadmin,
or an automated process.
[gmbricker@zanj Unreal3.2]$ ulimit -c unlimited

Bricker

2006-12-21 19:49

reporter   ~0012883

it was more or less, like i said syzop a kind of weird case. Very random and no apparent cause. feel free to close this if you'd like, I was just wondering if you've heard heard of anything similar to that

:)

Bricker

2006-12-27 20:28

reporter   ~0012923

OK, here is the same fsking error...lemme tell you what happened.


I upgraded the server, and relinked. everything was going good for a little less then an hour, then it delinked and I got this error


[Wed Dec 27 17:09:04 2006] - TIME SYNCH: Unable to synchronize time: Timeout. This happens sometimes, no error on your part.


so it didnt crash, just delinked...which was odd! wtf!

djGrrr

2006-12-28 08:45

reporter   ~0012924

Last edited: 2006-12-28 08:46

turn off time synchronization, its known to cause problems in some cases

make sure the system clock is synced regularly

syzop

2006-12-28 09:25

administrator   ~0012925

time synchronization is only done the first couple of seconds during booting (before the ircd is doing anything else), so it's not possible to get a timesynch error in whatever situation when the ircd is already up. (it is, of course, possibly that you tried to boot the ircd [a 2nd one] again at that time).

Time synchronization is not known to cause any problem.

djGrrr

2006-12-28 09:31

reporter   ~0012926

yes it is sysop :P

this is not the first time i've seen issues with the time sync

there was a thread over on anope forums a few days ago about it, something is definitely not only happening on statup
http://forum.anope.org/viewthread.php?tid=1141

but i think i see what he was saying, that error was coming from startup after the crash, which is normal, as i've never seen the default time servers ever work :P

syzop

2006-12-28 09:41

administrator   ~0012927

That's completely bogus, timesynch is only run the first 3 (max 5) seconds, and then never again. It is also run in it's own loop when nothing else runs.

You probably don't have too much experience with this yet, but people have bogus claims all the time... These are the same people that think moving the mouse in circles tends to speed up an installation process......... :P

Like Bricker's last comment is laughable if you understand anything from the code, but I don't say this... since he's a user and he cannot know :).

Same for katsklaw saying that Unreal has an "NTP server". What a joke.

Not to mention in that thread that VolTekka says he's getting bufferpool errors, a far worse error than anything else.

Also the default servers work fine, but several people (many shell providers) block UDP packets and such. I can't help that.

Ignorance is the right word here, and I don't mind it, as long as people don't persist with their idiotic claims :P

syzop

2006-12-28 09:42

administrator   ~0012928

We cannot help in any way without a core file or backtrace. Closing bug.

Issue History

Date Modified Username Field Change
2006-12-20 17:51 Bricker New Issue
2006-12-21 02:48 stskeeps Note Added: 0012869
2006-12-21 07:43 syzop Note Added: 0012871
2006-12-21 07:44 syzop Note Edited: 0012871
2006-12-21 10:29 Bricker Note Added: 0012876
2006-12-21 19:49 Bricker Note Added: 0012883
2006-12-27 20:28 Bricker Note Added: 0012923
2006-12-28 08:45 djGrrr Note Added: 0012924
2006-12-28 08:46 djGrrr Note Edited: 0012924
2006-12-28 09:25 syzop Note Added: 0012925
2006-12-28 09:31 djGrrr Note Added: 0012926
2006-12-28 09:41 syzop Note Added: 0012927
2006-12-28 09:42 syzop Status new => closed
2006-12-28 09:42 syzop Note Added: 0012928