-
Notifications
You must be signed in to change notification settings - Fork 575
Once started Zalenium goes into a loop of tearing down a container, then starting a new one, without ever reaching the number of --desiredContainers
.
#624
Comments
How much memory does your machine have? 32 containers seems like quite a lot given you’re not using kubernetes. If you make the desiredContainers 10 does it work? I suspect if you want the scale you’re after you’re going to need to switch to kubernetes. When I run zalenium in kubernetes I assign 1Gb per selenium node and 1/2 a cpu which for you would be 32 GB ram and 16 cores. Which is a pretty large machine. And that’s ignoring zalenium itself and anything else running on the server. |
It's an Azure instance, just running Zalenium... Standard DS5 v2 (16 vcpus, 56 GB memory) It seems to be a recent update to Zalenium that's caused this error (I had to go back to docker-selenium for a while). I can try rolling back to a previous version and report back. I do need to look into Kubernetes in Azure and VSTS. |
@matthew-horrocks what version of zalenium were you running beforehand? @bramvdklinkenberg is running Zalenium on the Azure Kubernetes Service (AKS) after a little bit of effort, see #603 for some of the detail. |
I have the same issue |
Hi, sorry, I've been away and haven't had time to look at this. I'll roll back the version of zalenium tomorrow and find out when the problem started. |
What's the best way to roll back to a previous Zalenium version? Change the How do I find out what |
nudge @pearj I have free time for this over the next couple of days. |
@matthew-horrocks yes you can roll back to previous versions that way. Generally, it's a good idea to keep the same selenium version, so for |
@pearj finally nailed this to a release. This bug first appeared in https://github.com/zalando/zalenium/releases/tag/3.11.0f If I roll back to 3.11.0e, the bug goes away and I can start and use all the 32 containers. |
Ignore this comment. I had a brain fart :) |
3.11.0f was a fairly significant rewrite of how zalenium starts containers. @dspasojevic did most of the work here, he might have some insight into your problem. |
Looking at the first set of logs, I see the first container of
But going missing from the point of view of the docker container a few seconds later:
Zalenium then attempts to start a new container because
I don't see anything in the first set of logs that indicates why the containers are being terminated. It would be helpful to see the shutdown reason for those containers if you can get it from your docker instance. I'll look through the debug logs a little later. |
@matthew-horrocks do you see the same behaviour when running with - for example - 10 desired containers? |
@dspasojevic |
Thanks. Can you see termination reasons for the containers from docker's point of view? |
What's the best way to get those logs? If a container has been terminated, aren't the logs lost? |
maybe |
Or it might be in your system logs? |
Sorry for the delay getting this info... broke my collar bone a week ago, so out of action for a while :( got some logs from a failed container...
|
I have started docker in debug mode with
|
I was trying to reproduce by starting an EC2 instance with Ubuntu, 16CPUs and 64GB RAM. I was able to create the 32 containers without any problem, but what I saw is that the logging is misleading, since the I will push a change that should help this issue, and if it does not fix it, at least we will be able to diagnose it better. |
in case --desiredContainers is bigger. Should fix #624
) in case --desiredContainers is bigger. Should fix #624
Just merged this into master, will release today. So please let us know how it evolves after the release, feel free to reopen the issue in case it does not solve it. |
Hi @diemol I just tried this with 3.14.0c and the same failure occurs... I turned on debug logging and found
I don't know if the If I set It seems like this should be re-opened. |
This now seems inconclusive as the number of |
Same here. Why this bug is closed? |
Nudge @diemol :) |
still there on latest release |
We'll reopen and check again. |
Hi all, Could you please try again with the latest release? Different improvements in the internals have been done, I tried myself to create 32 containers and it went well. So we would like to get some feedback from the latest release https://github.com/zalando/zalenium/releases/tag/3.141.59d |
Hi, I've tried again with https://github.com/zalando/zalenium/releases/tag/3.141.59d and found that the containers are continually being torn down and recreated. Please also see bug #786 it appears they are also having the same issue with 3.141.59d |
Would it be possible to see the log please? |
@diemol just tried with 3.141.59d zalenium container log
Started with:
Running on MacOS Mojave 10.14 |
@straris it is a bit confusing, the issue is about not being able to start 32 containers properly, and I kind of see in your log that something goes wrong while running tests. So maybe you can open a new issue describing the problem you have? It is hard for me to understand what is going on when different things get mixed under the same issue, but in the end the logs are not showing the same issue. |
@diemol my mistake, I assumed the issues are connected due to the same exception. |
No worries @straris :) |
Logs from a conversation in Slack @matthew-horrocks would it be possible to have the logs you got after trying the latest release? |
Logs for latest release https://gist.github.com/IGenegabus/aa68617103584e76ce4d67ae5eb12d60 |
* Avoiding the removal part since we don't use the same proxy in a long period of time, and it will likely be removed when the proxy shuts down. * Checking if the proxy has already been registered in a previous request, maybe the current thread is a subsequent registration attempt from the proxy. This fixes #624
I was finally able to reproduce this during the weekend, and I pushed a commit that fixes it, which is part of #793. |
Great, I'll keep an eye out for the new release. |
This was fixed in https://github.com/zalando/zalenium/releases/tag/3.141.59e |
Zalenium Image Version(s):
Docker Version:
Docker version 17.12.1-ce, build 7390fc6qq
If using docker-compose, version:
OS:
Docker Command to start Zalenium:
sudo docker-compose up --force-recreate
Expected Behavior -
Zalenium starts all requested containers.
Actual Behavior -
Once started Zalenium starts a number of containers, then goes into a loop of tearing down a container, then starting a new one, without ever reaching the number of
--desiredContainers
.The text was updated successfully, but these errors were encountered: