New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Memory leak in Reactor #37938
Comments
@DmitryKuzmenko Can you take a look at this? I was able to confirm the memory leak with just the salt-api webhook triggering highstates on minions. |
I can't say anything without a deep analyze. Now I'm working on another memleak issue. I'll add this one to my ToDo list. |
@DmitryKuzmenko yes, let's make this a top priority since it is a blocker for an upcoming release. |
Same behaviour on 2016.11.0 (Carbon) |
@johje349 I am having trouble replicating this to get to over 1G of memory. Are you sure that your ps command shows the command at 10G of Residential memory? I believe ps returns the RES number back in KiB, which would mean that you are at just over 1GiB of memory in the ps command in your first comment. Also, the percentage looks low, the Mem percentage looks like it is 6.6%, which would mean you have about 15-16G of ram in the server. Are you sure you are seeing this go up to 10x this amount of ram? I have a docker-compose environment simulating spinning up and down, but it is going to take a while today to get up to 1G. Thanks, |
@johje349 I think this may be a leak in the RunnerClient. Would you be able to convert your orchestrate script over to just a reactor sls that runs Thanks |
@gtmanfred I have 16G RAM in the in the server. Currently 13G - 6G Cached is used.
I will test your suggestions. BR, |
ZD-1056 |
@gtmanfred Currently I have only beacon load events coming in, which triggers a Reactor that runs local.state.sls. These are small events but I can see memory usage is increasing with about 4k per event. It would take quite a while to reach 1G+ with this configuration. BR, /JJ |
I have now an uptime of 20 days, and the Reactor process has consumed more than 2G of memory: |
@johje349 I'm working on this issue. It's actually not as easy. |
@DmitryKuzmenko change applied, let's wait and see. |
The fix does not seem to have improved the situation unfortunately. |
@johje349 bad news. =( This means I've reproduced not yours issue. Sorry. |
@DmitryKuzmenko salt-master has been running for 4 days now and the Reactor process consumes 800M, it seems to be pretty stable around that value though. |
@johje349 glad to hear it. Could we close this issue then? |
@DmitryKuzmenko yes go ahead and close it. |
@johje349 thank you! |
Description of Issue/Question
Memory leak, most notable in Reactor and EventPublisher. The Reactor process has over 1G reserved memory. The leak seems to occur during orchestration following minion creation using salt-cloud.
Total used memory after a day of uptime has gone from 2G to 10G.
Possibly due to queuing the states in the orchestration. There are between 70 to 100 minions in the setup (we do auto scaling).
Setup
A reactor which fires on salt/cloud/*/created event:
The orchestration SLS:
Steps to Reproduce Issue
Create a few minions using salt-cloud with the above setup.
Versions Report
The text was updated successfully, but these errors were encountered: