Page MenuHomePhabricator

Deploy separate anonymous session backend to Wikimedia production
Open, Needs TriagePublic

Description

Set the anonymous session store to whatever was decided in T402850: Decide on anonymous session backend. By this point, we should have verified via T402808: Deploy separate anonymous session backend to Wikimedia production, in log-only mode that the split backend works as expected, so this is low-risk (unless there's risk on the infrastructure side). We can probably just roll to group 0/1/2 in three successive days.

When a wiki switches, all current anonymous sessions are lost. That should be fine, a few people will have to restart login/signup or resubmit their edits but there will be no data loss.

Event Timeline

Tgr added a project: OKR-Work.
Tgr added a project: User-notice.

Probably worth a Tech News note, it shouldn't affect editors in any visible way but it's interesting.

@Tgr how can we word this? Asking for Tech News.

Per T402850#11246764 the anonymous cache will also be Cassandra / Kask, but a separate instance. @Eevans please let us know when you are ready to proceed.

Is the relevant task T392170: sessionstorage namespacing? Also, do you still want the splits mentioned there (per wiki or deploy group)? That would be an easy addition if you think it is useful.

@Tgr how can we word this? Asking for Tech News.

Not sure if it's worth adding, it won't be observable to editors (unless something goes very wrong).

If you want to mention it anyway, maybe something like

A session store is a database used to remember users of a website between successive page loads. Wikimedia is splitting its session store into separate instances for anonymous and logged-in users, to reduce the risk of web scrapers interfering with logged-in users' usage of the sites. There should be no visible changes.

or if that's too long,

A separate session store has been created for anonymous users, to reduce the risk of outages from web scrapers. There should be no visible changes.

I like that last one, but I'd flip it around. First the result, then the technology.

To reduce the risk of outages from web scrapers, a separate session store has been created for anonymous users. There should be no visible changes.

Thanks!

@Tgr how can we word this? Asking for Tech News.

Not sure if it's worth adding, it won't be observable to editors (unless something goes very wrong).

If you want to mention it anyway, maybe something like

A session store is a database used to remember users of a website between successive page loads. Wikimedia is splitting its session store into separate instances for anonymous and logged-in users, to reduce the risk of web scrapers interfering with logged-in users' usage of the sites. There should be no visible changes.

or if that's too long,

A separate session store has been created for anonymous users, to reduce the risk of outages from web scrapers. There should be no visible changes.

I like that last one, but I'd flip it around. First the result, then the technology.

To reduce the risk of outages from web scrapers, a separate session store has been created for anonymous users. There should be no visible changes.

Thanks!

@STei-WMF: Please avoid unneeded full quotes - keeps a task easier to read. Thanks. :)

[ ... ]
Is the relevant task T392170: sessionstorage namespacing?

Let's use that as the tracking issue; I'll create subtasks of T392170 as necessary.

Also, do you still want the splits mentioned there (per wiki or deploy group)? That would be an easy addition if you think it is useful.

For now, we should probably stick with anon sessions. I will open a subtask for further splits though so we can discuss what that would look like/entail.

Not sure if it's worth adding, it won't be observable to editors (unless something goes very wrong).

I will remove the #user-notice tag then, and unsubscribe. If you reconsider, please re-add the tag, or add the draft to the relevant edition of Tech News at the right time. Thanks.