Tue, Aug 20
Here are the creation statements:
Mon, Aug 19
There already is a mechanism in change propagation to back off and wait / retry later. We use this when MW is set to read-only. If the cirrus jobs could indicate that to the job runner (e.g. by raising a ESReadOnlyModeException or something similar), it's easy to make it understand that.
Thank you, @RiskNerd for reporting, investigating and fixing the bug :)
Yay \o/, really appreciate it @fgiunchedi !
Wed, Aug 14
Moving back to CPT's backlog since the PS' are WIPs. @matmarex feel free to ping us when you feel the patches are ready to be reviewed.
Tue, Aug 13
I have looked into this, but couldn't find an exact root cause. To me this smells like a cache data corruption issue. Here's some data:
Perfect, thank you @MSantos
Mon, Aug 12
Fri, Aug 9
I went through the patch and it looks like it DRT, but I'd like @Krinkle to give it another pass before we merge it.
Thu, Aug 8
This seems to be a wmf.16 regression (as this error is not appearing for earlier deploys).
Wed, Aug 7
The branch has already been cut. so it's not relevant for this week's train.
Fri, Aug 2
We're not using k-r-v any more, resolving.
Jul 5 2019
Jun 27 2019
This is actually Citoid reporting a 404 for that website. I haven't looked to deep in it, but it seems that the website has troubles producing something meaningful if not retrieved by a browser. @Mvolz could you look into this?
Regarding the deployment plan, the main pain point is that we will need to have both front-end and back-end processes behind the same LVS and then slowly withdraw the back-end ones behind a separate LVS end point. As doing so while running all of them on the same port poses some challenges, @Joe and I came up with an interesting idea. For the case of the back-end service, we could have it bind to two different ports: the one eventually to be used by the front-end service and the one to be used by the back-end service. That way, we can shift processes from one LVS to the other seamlessly. On the implementation side, this can be achieved by levering service-runner's multiple services inside one process functionality: essentially, we can declare the back-end service twice in config.yaml, but assign it different ports:
This is ... DONEEEEE \o/