How much memory does MySQL Enterprise Monitor’s agent use?
After last week’s post on agents versus agentless monitoring systems, I got a lot of email. One, from a customer whose name I am not permitted to mention, sent me the following action shot (posted with permission):
Over half a gigabyte; more than twice what MySQL itself is using. So, that raises an interesting question. How much memory would you say your server’s MySQL Enterprise Monitor agent uses? No bashing allowed, tell the truth. And please post if your agent runs in a small amount of memory, too.
Another thing I’m interested in: what factors contribute to the high memory usage? Under what conditions does it use a lot or a little? What can you expect, generally speaking?
On another note, I think it’s almost unfair not to mention the competition, so I will: I was playing with MONyog 2.6 recently and it started to use a lot of CPU. I think it might be architecture-specific (amd64 Ubuntu) but I can’t be sure. It was using 170% CPU on my dual-core system to monitor 2 instances.
Agent memory has never been an observed issue for me. I just checked a dozen of my production servers (All Ubuntu Linux). Almost every single one was taking up 300K within 10% either way. Didn’t seem to correlate with how long the agent had been running or how much activity the server gets — however the longest they had been running is about 3 weeks when I upgraded to the latest stable version.
The 2.0 beta is taking up 370K on one of my QA servers while running the query analyzer module on top of the current agent. However that has been running only 2 weeks, and not a lot of constant activity. Still, not something I’m really concerned about. The bottleneck for me has always been the centralized server which aggregates data and displays the charts/graphs.
Ryan Thiessen
26 Aug 08 at 5:38 pm
Ryan, are you certain your KBs aren’t MBs? Just to be sure…
Your numbers are about 1000x off from Barons, AND it would be quite rare for any tool these days to use less than a meg or more.
Arjen Lentz
26 Aug 08 at 6:11 pm
Ours runs about 380 MB on Linux (on either heavy use or low use servers).
We did move the centralized database off to it’s own database server which helps the graphs/cpu/memory usage on the enterprise manager console server.
Phil Hildebrand
26 Aug 08 at 6:18 pm
It would be interesting to note which OS the agent is running on, the agent version the elapsed time the agent has been running, and how many advisory rules are schedules on that agent.
Agent 1.3.2.9118
$ uname -r
Linux 2.6.24-19-generic
# ps -wo pid,etime,rssize,vsize,comm `pidof mysql-service-agent`
PID ELAPSED RSS VSZ COMMAND
10707 44:46 4020 74088 mysql-service-a
Full load of 98 rules and 20 graphs.
Also, There were some fixes in the most recent 1.3 agent to reduce the memory footprint for agents monitoring large instances (many dbs/tables).
Matthew Montgomery
26 Aug 08 at 7:26 pm
If I wasn’t clear on my agent memory usage:
Resident (physical) Mem = 4M
Virtual Mem = 73M
Matthew Montgomery
26 Aug 08 at 7:33 pm
Arjen — yes of course that was M not K. Sorry for the confusion…
Ryan Thiessen
26 Aug 08 at 8:48 pm
To be fair, MONyog is a centralized server, not an agent, so having it use a bunch of CPU isn’t unexpected….have you compared it to MySQL’s Enterprise Monitoring *server*’s CPU usage?
Sheeri K. Cabral
26 Aug 08 at 10:04 pm
Baron,
We have been trying very hard to reproduce the situation when MONyog consumes 170% memory, but no success till now :-(
We are trying on Ubuntu, but we don’t have amd64 hardware available at our offices. We will get one today!
Anyone else experiencing high CPU issue with MONyog?
Rohit Nadhani (Webyog)
26 Aug 08 at 11:30 pm
The memory usage of the agent should be around 10MByte. If it is more, please report a bug at http://bugs.mysql.com/
Jan Kneschke
27 Aug 08 at 5:20 am
On two linux x86_64 replicas, VSZ is around 405M and RSS is 135M, give or take a few megs, and CPU usage is very low (
Devananda van der Veen
28 Aug 08 at 12:07 am
For us, running on Solaris and with QA on (for 3 days now), one of our agent is using 1,1 GB of memory now… We’re currently investigating the issue with Oracle (it’s been on for a while and happens only in one MySQL node, not with the others).
At one time, when the Agent was using about 800 megs of memory the queries started to be slow -> we had to restart the Agent. That was not cool…
I
17 Feb 12 at 5:54 pm
there is clearly SLOW memory leakage on agent.
In my case, memory starts from 300 MB to 900 MB
over 4 weeks.
I now have to restart agent once a week
jimmy chen
11 Mar 13 at 5:51 pm