Ad Widget

**Palmertree** · 10-08-2007, 05:57

I was just about to post the same problem. I have to restart zabbix_server ounce a day because the memory for the zabbix_server process slowly increases by the day.

**Saftnase** · 13-08-2007, 10:45

dito can confirm this behavior, but need to restart my zabbix server only every 5 days

**JonB** · 13-08-2007, 12:05

I also had the same issue with 1.4.1 on 2 CentOS 4.3 based servers. In both cases the agents had been compiled with the --static option.

Once I recompiled without the --static option the problem went away.

Jon

**Palmertree** · 13-08-2007, 19:53

Interesting....

I wonder why the --static option has an affect on the memory consumption.

**Alexei** · 13-08-2007, 21:24

Originally posted by Palmertree

Interesting....

I wonder why the --static option has an affect on the memory consumption.

It sounds like a memory leak in a third party static library, not in ZABBIX code itself.

**Palmertree** · 14-08-2007, 06:20

Are there any good linux utilites out there to help pin point the problem?

**Alexei** · 14-08-2007, 07:52

Originally posted by Palmertree

Are there any good linux utilites out there to help pin point the problem?

Hmm, ZABBIX?

I would be interested to see ZABBIX server memory usage for a period of time, not usage of swap space.

**drose12** · 14-08-2007, 17:25

Originally posted by Alexei

Hmm, ZABBIX?

I would be interested to see ZABBIX server memory usage for a period of time, not usage of swap space.

The swap space use graph was the results of the zabbix processes taking up all the memory. The funny thing on my system is that I have 36 zabbix_server processes, but not all of them are consuming memory. And the parent process looks fine. Without knowing which pid to watch, how can I go about graphing the memory usage of 35 processes? I have 10 right now which are 185mb each right now. Looks like I require a restart. BTW I did not compile my server with the static option.

**bbrendon** · 15-08-2007, 08:23

I think I'm leaking too. Not sure yet, but I found this.

http://www.pixelbeat.org/scripts/ps_mem.py

Interesting.

ps_mem.py reports:
224.0 MiB + 2.3 MiB = 226.3 MiB zabbix_server (20)

proc.mem[zabbix_server] reports: 1,342,836,736

1.3 Gigabytes vs 226 megabytes ?!??!?

WOW.

**pixelbeat** · 15-08-2007, 16:51

Originally posted by infinity005

I think I'm leaking too. Not sure yet, but I found this.

http://www.pixelbeat.org/scripts/ps_mem.py

Interesting.

ps_mem.py reports:
224.0 MiB + 2.3 MiB = 226.3 MiB zabbix_server (20)

proc.mem[zabbix_server] reports: 1,342,836,736

1.3 Gigabytes vs 226 megabytes ?!??!?

WOW.

Note ps_mem.py reports the RAM used.
I.E. any memory swapped out to disk is not reported.

In summary ps_mem.py reports the current amount of
RAM used per program, rather than the current amount
of memory used per process, as reported by traditional tools

**bbrendon** · 15-08-2007, 18:57

Whoa. Did I just take that script from your website? And you're here?

Creepy. That was un-intentional.

...

Wait. You have 1 post! Are you monitoring your weblogs in realtime? Are you a stalker?

Please brain dump on us with all your knowledge while you're stalking. Thanks!

**bbrendon** · 15-08-2007, 19:32

Enough fun. Back to the topic at hand.
My zabbix server has a LOT of stuff running on it. So hopefully others can help with this.

I created a proc.mem[zabbix_server], and so far (12 hrs) it has increased from 1342836736 to 1376583680

Difference: +33,746,944

I'm running Debian etch, amd64.

How can we find the library causing the leaks? I don't think I have enough data to say zabbix is leaking, but based on my swap data collection and the time I'm restarting the zabbix_server, it seems zabbix IS leaking.

**drose12** · 15-08-2007, 19:53

Originally posted by infinity005

Enough fun. Back to the topic at hand.
My zabbix server has a LOT of stuff running on it. So hopefully others can help with this.

I created a proc.mem[zabbix_server], and so far (12 hrs) it has increased from 1342836736 to 1376583680

I've created mine as well...I'll post my results after a day or so as I've just restarted zabbix to show the trend.

**bbrendon** · 16-08-2007, 19:06

So this is getting to be a problem. I though our server was going to crash and burn today. Luckily it came back after about 5 mins. When it was finally responding, i typed "uptime"

load average: 44.41, 29.11, 14.10

So it appears a simple stop of zabbix caused the load to go from a normal 1-2 to 44

Attaching my proc.mem[zabbix_server]

Help!

Attached Files

Ad Widget

How to tell if Zabbix has memory leak?

How to tell if Zabbix has memory leak?

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment