copied just the config of list "test" from sodium and put it in place on fermium

first there was a mailman bug page but the issues were just file system permissions

(also check the "en" subdirectories and similar, multi-language listinfo pages!)

then we ran withlist -l -r fix_url test -v which fixes the URLs inside the .pck config files and adjusts them to the current server (fermium)

additionally we used mmsitepass to reset the site password, literally _re_set to the same password it was before but it is necessary to be able to login on the imported list using the site pass. the regular list admin pass works without changes

after setting the list to advertised it also showed up on the global listinfo overview

Is this done in theory or is there something else needed before this can be marked as 'done'?

Well, originally we thought "export" might mean more than just rsync. Now we know it doesn't.

Technically it is only done when we import it the second time on the server that will actually be production though.

for some reason the "reset password" thing above wasn't needed in all cases. at least in one case i could login with the site password without re-setting it and after freshly importing a list

running rsync of all archives to new fermium, with --dry-run

Change 233888 had a related patch set uploaded (by Dzahn):
mailman: mini script to rsync lists

https://gerrit.wikimedia.org/r/233888

Change 233888 merged by Dzahn:
mailman: mini script to rsync lists

https://gerrit.wikimedia.org/r/233888

Change 233974 had a related patch set uploaded (by Dzahn):
mailman: adjust rsync script, exclude bounces

https://gerrit.wikimedia.org/r/233974

sent 4397116660 bytes received 1973579 bytes 2133926.87 bytes/sec
total size is 4389356412 speedup is 1.00
done rsyncing\n

real 721m16.384s
user 6m11.660s
sys 11m33.910s

for an initial run plus a little more, maybe another half hour that i had already been running a sync for just archives

and doing ./data/ one more time after adjustments to the rsync command

Change 233974 merged by Dzahn:
mailman: adjust rsync script, exclude bounces

https://gerrit.wikimedia.org/r/233974

plus about another half hour to re-sync the heldmsg-* files (which used to be >500k but now are just over 100k

repeating a run of ./rsync_lists.sh right after it was finished takes:

63m23s

so it takes a minimum of an hour to run but actually more depending on how much time there is between rsync 1 and rsync 2

deleted all files in qfiles/bad , tons of files and we don't use them

also patch by John https://gerrit.wikimedia.org/r/#/c/234043/4 to stop them from coming back

next is checking qfiles/shunt which is another 2.5GB , cronjobs should be running that delete them after 7 days, but don't

deleted all files in shunt older than 7 days, also see T110382

even after removing all this stuff, an rsync (second run) still took 60m, exactly an hour

Change 237287 had a related patch set uploaded (by Dzahn):
mailman: change the way we rsync stuff

https://gerrit.wikimedia.org/r/237287

Change 237287 merged by Dzahn:
mailman: change the way we rsync stuff

https://gerrit.wikimedia.org/r/237287