-
Notifications
You must be signed in to change notification settings - Fork 504
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
Linkerd CPU usage climbs over time #2268
Comments
Thanks for filing this issue @yeyincai. Could you tell us a little bit more about the "IO connections" increase? Are these connections to clients or servers that are still active? Does that number fluctuate and still rise in an upward trend like the CPU graph above? I wonder if those "EmptyLastHttpContent" corresponding with connections being left open. |
@dadjeibaah Thanks for you , we have not IO connections graph, at that time view with cmd= "netstat -ano |grep 4140". latency increased ,a lot of request time_out ,may be the connections is closed. |
@adleong This problem is still bothering us now, usually restart,i hope you can help solve,thanks!We find the connection is not closed,we use iptable partition network,but incomming connections is not closed . We use linkerd as httpgateway, I don't know if it will affect |
@yeyincai, do you have any info about the clients that are connecting to Linkerd? That incoming connection count seems a little peculiar but it may be fine depending on the behavior of the clients Linkerd is connected. I have a couple of questions:
|
Hi @dadjeibaah, it seems like we have fixed the issue by adding "-XX:-UseBiasedLocking" and so far it works fine. The idea was from the safepoint log we saw a lot of "RevokeBias". Have you guys tested about the performance impact of enabling/disabling biased locking? I think there should be a reason for keeping "UseBiasedLocking" in default, right? |
@Neron-Liu yes ,final we adding "-XX:-UseBiasedLocking" ,the problem is solved |
Hi @dadjeibaah I work with @yeyincai, to Q1: Linkered runs in our internal network, so it is dedicated for internal traffic. to Q2: The client is Nginx, and we use HTTP/1.1. The origin traffic is from PHP. to Q3: I just go through to the tcpdump files, no bad TCP state was found. We are going to reproduce the issue, will update you later. Pls let me know if you need any information to get this solved. Thanks. |
Thanks for the great investigation and report @yeyincai and @Neron-Liu! What would you like the next steps to be here? Have all your problems been resolved by the use of the |
Issue Type:
What happened:
We use Linkerd with http1 protocol ,Linkerd CPU usage climbs over time ,and IO connections counts climbs over time ,GC count climbs , after run 7 days , are a lot of exceptions,but we restart Linkerd it was normal .
jstack dump:
cpu:
linkerd.log:
Environment:
The exception "java.lang.IllegalArgumentException: invalid message 'EmptyLastHttpContent'",we suspect of that Finagle HttpDecoder has a bug.See the new finagle-19.4.0-release, They modified finagle-http with Http Chunked.
The text was updated successfully, but these errors were encountered: