1.4 linux server keeps crashing

[Buffalo]

Hello,

Server is running debian-6.0, MTA 1.4-release-6760, using pre-compiled server files from linux.multitheftauto.com.

Error on crash:

http://i.gyazo.com/6c572078cf1f88abf687e7119a6b3d98.png (bad_allocation?)

Crash happens after awhile, somewhere in the middle of 20mins to 4 hours.

Gamemode scripts are mostly running client-side. Script is based on peds.

Major server scripts based on custom files downloading, that works very well. E.g. no open file handles reported by performancebrowser. Actually can't see anything abnormal in performancebrowser.

Although, there are some server side peds, like bosses, that must be synced correctly.

Also, this gamemode is 1 year old, with nothing special added (except file downloader), it worked perfect on MTA 1.3.3.

I forgot how to debug crash deeper, since I had last server crashes on this gamemode on early MTA versions. Once someone reminds what to use, I will try to collect more information it dumps.

EDIT: Crash seems pretty random. Worked 9 hours straight without incident.

Edited August 19, 2014 by Guest

[kidvnet]

I have same experience. what I did is to read the dump file. even it was encrypted. finally I found some information just above the hardware description. there is a resource name, which was a dayZ addon for anti-corpse-copy. I guess maybe some variables or data table overflow. so later I write a script to restart that resource every half hour. now the server is never crash

[Buffalo]

At least you can read a dump file. I never got one.

BTW nice english.

Edited August 19, 2014 by Guest

[ccw]

bad_alloc can be caused by the process running out of memory, but it could be something else. Is there a dump file in mods/deathmatch/dumps ? If so, upload to http://upload.mtasa.com/

If dumps are not being generated, here is some info on manually getting a back trace:

https://wiki.multitheftauto.com/wiki/In ... er_crashes

.

[Buffalo]

Right, so heres my homework:

http://pastebin.com/sjQterXk

(There was no core dump file)

Regarding this info, im using latent events to send files. It works good on other server.

[ccw]

Thanks.

I'm almost certain the crash is due to the process running out of memory. Can you PM me the script that does the latent sends to the clients?

[Buffalo]

Any information how to stop this would be appreciated. What exactly could cause these leaks (i guess?). Or if it's MTA related issue, I'll wait for a server patch version to update to.

EDIT: Last crash had different error: http://i.gyazo.com/eaecaa23912af6ad55523700bf3af5c3.png

[ccw]

Try the latest nightly build from here:

https://nightly.multitheftauto.com/

[Buffalo]

Uploaded linux-1.4.0-rc-6823

Still crashes with same error (bad_alloc).

[ccw]

Change your triggerLatentClientEvent code so the server only sends one file at a time - This will prevent the server from queuing all the file requests (which uses memory).

Also make sure the server never sends the same file more than once for any client - This will prevent hackers causing memory problems by constantly triggering the 'download' events

[Buffalo]

Thank you for support. Rewritten scripts as you advised, crashes are gone.