-
Notifications
You must be signed in to change notification settings - Fork 4.9k
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
Kong2.3.0, cpu 100%, nginx No response #6739
Comments
@RyouZhang, please give us more information. Did you do inplace update? If so, you might have conflicting rockspecs installed. |
Use docker deploy after docker restart 24h, kong cpu 100%, and not response; It's very interesting. I do a case, one kong has normal traffic,another only has metric; but after 24h,cpu 100% and no response all of. I think maybe metric has some trouble. After 24h, the issue reappear,kong reload, can response but cpu 100% and memory leak happend Proc info strack -d : |
Having the same issue. Reproduction steps below: Follow the below execution flow:
kong.yaml:
error logs: |
Both DP and CP logs in debug level start filling up extremely quickly with
To the point where I can't even attach the error log in debug level neither from the DP nor CP because of how big in size they get after just a few seconds. |
@murillopaula thank you for pointers. It looks like a bug in healthchecks. @RyouZhang, I will look what is happening, and will report back. It seems like it is healthchecks related and happens mostly on hybrid mode. @RyouZhang what mode did you run yours? |
@RyouZhang do you use hybrid mode? do you use upstreams/targets with health checks (passive/active)? |
ok, this is my kong.yaml and docker cmd |
@RyouZhang ok, it seems like you are also using health checks. We've found a temporary workaround to mitigate this issue:
i.e. downgrade the Can you let us know if that works for you? Note this is a temporary hack and should not be considered a definitive solution. |
use lua-resty-dns-client 5.1.1, work 24h, it's fine. |
### Summary @RyouZhang reported with #6739 that Kong starts to spin CPUs 100% with `2.3.0` when healthchecks are enabled. @murillopaula confirmed it and gave additional information and we could reproduce it easily. We identified an issue in lua-resty-dns-client 5.2.1 together with @Tieske and @locao, fixed it. This PR bumps the Kong dependency to the fixed version of the library. ### Issues Resolved Fix #6739
@RyouZhang we released |
### Summary @RyouZhang reported with #6739 that Kong starts to spin CPUs 100% with `2.3.0` when healthchecks are enabled. @murillopaula confirmed it and gave additional information and we could reproduce it easily. We identified an issue in lua-resty-dns-client 5.2.1 together with @Tieske and @locao, fixed it. This PR bumps the Kong dependency to the fixed version of the library. ### Issues Resolved Fix #6739
Summary
SUMMARY_GOES_HERE
Steps To Reproduce
Additional Details & Logs
$ kong version
)$ kong start --vv
)<KONG_PREFIX>/logs/error.log
)https://docs.konghq.com/latest/admin-api/#retrieve-node-information)
The text was updated successfully, but these errors were encountered: