Welcome to Tool Labs
Hello Edsu, welcome to the Tool Labs project! Your request for access was processed and you should be able to login now via ssh at
tools-login.wmflabs.org. To get started, check the help page and the migrating from the Toolserver page. You can also ask in our IRC channel at #wikimedia-labs connect or send an e-mail to our mailing list firstname.lastname@example.org – thank you, and have fun using Tools! Tim Landscheidt (talk) 07:01, 16 November 2013 (UTC)
Completing project documentation
This is regarding the announcement (http://lists.wikimedia.org/pipermail/labs-l/2014-July/002762.html) if you have missed it.
We have added some fields in addition to description field which is already there on your projects. Please click "Edit Documentation" in your project page/ project pages and fill the blanks. It will only take few minutes from your valuable time.
Please correct: congresseditors.js running on tools-login rather than job grid
/data/project/congressedits/node/bin/node /data/project/congressedits/congresseditors/congresseditors.js is running on a Toolforge bastion server (tools-bastion-03 aka tools-login) rather than on the job grid. This violates the "Do not run noninteractive processes on the bastion servers" rule. Non-interactive processes on the bastion servers can be killed at any time by an administrator when resources there are abused. Please read the job grid help documentation and move your bot to the job grid.
- Thanks. I've been trying to get the version of Node upgraded on GridEngine because there is a memory leak related to the version that is being run. https://phabricator.wikimedia.org/T166830 I will move it over to kubecongrnetes which has a more modern version of node. edsu (talk) 20:55, 8 November 2017 (UTC)
- We are not going to upgrade node globally on the job grid, but anything you have working locally on the bastion will work on the grid. The same directories are mounted on all of the grid exec hosts. Moving your bot to Kubernetes is probably a much better long term solution. Setting up a bot there is a bit more difficult that submitting a grid job, but it is definitely possible. See Help:Toolforge/Kubernetes#Kubernetes_continuous_jobs for an outline of how I setup a python3 bot on Kubernetes. I'd be glad to help you get your node bot running there as well. --BryanDavis (talk) 00:16, 9 November 2017 (UTC)
- Thanks for the offer. I've just attempted to get congressedits working under kubernetes and it seems to be working--or at least running. Hopefully the memory leak associated with the tools use of jdom will be resolved. Do you know how memory leaks will manifest with kubernetes? Also, the project is actually misnamed as congressedits when it should be congresseditors. Is it very involved to change the project name? edsu (talk) 01:46, 9 November 2017 (UTC)
Do you know how memory leaks will manifest with kubernetes?
- I imagine that the nodjs process will crash eventually. When that happens the container ("pod" in Kubernetes terms) should die. Then the Kubernetes reconciliation loop should notice that you have a Deployment that is supposed to have 1 running pod and start the pod back up. The problem in this would be if nodejs locks up instead of actually exiting when it hits the memory cap.
Is it very involved to change the project name?
- We do not have any way to rename an existing tool account (or actually to delete one), but making new tool accounts is inexpensive as far as server resources are concerned. You could create a new tool with the name you would like to have and move the code over to it. After you get things running under the new tool account you can create a child Phabricator task of our tracking task for tools to delete. "Someday" we will get around to making a tool deletion procedure. --BryanDavis (talk) 02:31, 9 November 2017 (UTC)