Trying to repair it manually:

REPAIR TABLE group2view USE_FRM;
...
The storage engine for the table doesn't support repair

See https://stackoverflow.com/questions/2428738/how-do-you-fix-a-mysql-incorrect-key-file-error-when-you-cant-repair-the-tabl

@Andrew Works again, thanks!

Yes. No joy.

Sorry, but I don't think it's done yet.

@rook Thanks, the project is called baglama2. At the moment the volume size shows 60GB, which I suggest to increast to 100GB.

It does not, unfortunately:

Fixed now.

Or put the tool name in there so I can see it's mine...

tl;dr: I was on vacation with no ssh access. Webservice restarted (which could/should be done by Toolforge automatically but isn't), seems to work again.

This is still happening, reproducibly, so not a stale server or something

Thanks. It would be helpful if that were mentioned in https://www.mediawiki.org/wiki/Help:Extension:WikibaseCirrusSearch

Works again

The Cloud-Services project tag is not intended to have any tasks. Please check the list on https://phabricator.wikimedia.org/project/profile/832/ and replace it with a more specific project tag to this task. Thanks!

Hi @fnegri if this will take a while, could you throw me a few more DB connections for mix-n-match on ToolsDB in the meantime? I have a lot of background tasks running to keep everything useful and in sync with Wikidata.

Thnaks, that was it. Could have a better error message but if the old host is shut down soon anyway...

Update: API response is {"code":"unexpected-error","message":"Unexpected error"} for the second call.

At least it should fail with an "edit conflict" error or something, but since the statements are different, I would expect both calls to succeed

I am currently re-running all wikis to fix a trivial bug that led some articles to be marked as done when they were not. The issue is closed on a code level.

yowiki does not have any candidates. The wiki drop-down did show all pages in the database, not the available ones. This now fixed. I will investigate why yowiki has no more candidates.

Thanks, deployed, please test with your usecase.

I have

• deleted petscan4
• deleted petscan4a
• created petscan5 (as debian-12.0-bookworm)
• created a new web proxy to point to petscan5 (only seens to offer https://petscan.wmcloud.org/ instead of the previous https://petscan.wmflabs.org/ but redirect seems to work fine)

AFAICT this can be closed now.

Update: Building new VM now

Hello, didn't see this until now. Happy to move to to a new VM. Can you provide (and throw in a bit more disk space please?), or should I do this myself?

Silly question: Why don't we host the Docker images in the gitlab repo container service at https://gitlab.wikimedia.org?

See also: https://www.docker.com/increase-rate-limits/

Thanks @fnegri I changed the ticket title. Seems like an oversight not to add PHP?

Also no php as user:

magnus@tools-bastion-13:~$ php
-bash: php: command not found

Figured it out

FWIW this is the mix-n-match tool, in case it matters.

This tool would not scale to today's Wikidata size

It did happen three times in a row before I reported it, just worked now. Maybe a github problem or a server not seeing internet?

This appears to be reolved, and replaced by the too many open files bug (another Phab ticket is open).

@dcaro thanks, done

Got it working. For future reference:

I have some updates. With php.ini as

extension_dir="/layers/fagiani_apt/apt/usr/lib/php/20210902"
extension=mysqlnd
extension=mysqli

it loads the mysql library. Still fails with some password error but that seems like something code-wise to debug.

For my own testing, is there a way to get a shell into the docker container? I have seen the docs for the webservice one, but that doesn't seem to fly for non-webservice ones?

I created a custom php.ini and use it in the command:

php8.1 -c /data/project/mix-n-match/mixnmatch_rs/php.ini /data/project/mix-n-match/scripts/person_dates/update_person_dates.php 48922

The php.ini:

extension_dir="/layers/fagiani_apt/apt/usr/lib/php/20210902"

I also added a whole bunch of packages to the Aptfile, just in case.

@taavi I added php-mysql to the Aptfile and rebuild the image, same error.

Thanks @taavi Is there a list of packages required for a toolforge-equivalent php setup?

Looks like the php setup is incomplete. Running php8.1 path/to/my/php/script dies with Fatal error: Uncaught Error: Class "mysqli" not found.

Would it be possible to add a symlink in $PATH somewhere? php is stable, php8.1 maybe less so

Ah, I found /layers/fagiani_apt/apt/usr/bin/php8.1, I guess it's just not symlinked anywhere?

I added an Aptfile as described and rebuild the image, but:

tools.mix-n-match@tools-sgebastion-10:~/mixnmatch_rs$ toolforge jobs run --mem 2000Mi --cpu 1 --wait --mount=all --image tool-mix-n-match/tool-mix-n-match:latest --command "php --version" single
ERROR: job 'single' failed:
+-------------+------------------------------------------------------------------+
| Job name:   | single                                                           |
+-------------+------------------------------------------------------------------+
| Command:    | launcher php --version                                           |
+-------------+------------------------------------------------------------------+
| Job type:   | one-off                                                          |
+-------------+------------------------------------------------------------------+
| Image:      | tool-mix-n-match/tool-mix-n-match:latest                         |
+-------------+------------------------------------------------------------------+
| File log:   | no                                                               |
+-------------+------------------------------------------------------------------+
| Output log: |                                                                  |
+-------------+------------------------------------------------------------------+
| Error log:  |                                                                  |
+-------------+------------------------------------------------------------------+
| Emails:     | none                                                             |
+-------------+------------------------------------------------------------------+
| Resources:  | mem: 2000Mi, cpu: 1                                              |
+-------------+------------------------------------------------------------------+
| Mounts:     | all                                                              |
+-------------+------------------------------------------------------------------+
| Retry:      | no                                                               |
+-------------+------------------------------------------------------------------+
| Status:     | Failed                                                           |
+-------------+------------------------------------------------------------------+
| Hints:      | Last run at 2024-03-12T14:23:38Z. Pod in 'Failed' phase. State   |
|             | 'terminated'. Reason 'Error'. Started at '2024-03-12T14:23:39Z'. |
|             | Finished at '2024-03-12T14:23:39Z'. Exit code '127'.             |
+-------------+------------------------------------------------------------------+

I also tried commands like "sh -c 'php --version'" but no success.

Yes, the third attempt passed. However, 500 errors should be cause for concern, even if it eventually runs through.

Now happening repeatedly for the mix-n-match tool:

toolforge build start https://github.com/magnusmanske/mixnmatch_rs/

this has resolved itself

Yes, works for me, thanks

Ran it again now, works fine. Closing this issue, will re-open if it keeps happening.

Now starting to see error 429 (too many requests), which is better (for debugging). Did something change in the meantime?

This happens ~50% of the time, re-running the exact same command often works.

Thanks, that works for me. It is much slower than before, as it needs to run full compilation every time, instead of just the changed bits. Ah well.

I tried that and got

[step-results] 2024-02-21T11:49:25.959261833Z Built image tools-harbor.wmcloud.org/tool-listeria/tool-listeria:latest@sha256:24ac86ba3d8e4c11568e158a172016dcf9b8e697312866c674ad5ae6f6b4954b

But the tutorial shows me only how to deploy a webservice. How do I start the bot executable as a normal toolforge job?

Ah, gridengine doesn't compile it either... Works fine locally though

Seems fixed now

The Cloud-Services project tag is not intended to have any tasks. Please check the list on https://phabricator.wikimedia.org/project/profile/832/ and replace it with a more specific project tag to this task. Thanks!

The funny thing is, if you click on the movie https://commons.wikimedia.org/wiki/File:The_Sea_Beast_(1926).webm to play it, the moment the player comes up, it shows the correct duration. Where does it get that from?

I have limited it to 50 API connection at a time. Still throwing 104 errors. Either I did it wrong, or there is some other issue.

Thanks to the recently relaxed limitations per job on Toolforge, I am running this with more async threads, so without actual code change, it would have increased the number of connections.

Please use the bot as you like for testing!

This issue is persisting for the listeria tool. The bot is now down from ~40K edits/day to ~500. Something needs to be done, soon.

I am open to this, as I already have a Trove instance for baglama2, and it works well (after some initial problems).

FWIW I just saw this ticket, and added an index to the log table which should speed things up. I can also rewrite the query if that does not suffice, please let me know. Let's discuss Trove in T350862 .

D'oh!

Webservice is not running, returns only blank page, no error

Seems to be fixed now?

Aaaaand its back again.