Our ipsilon / fedora openid provider is being very slow to respond.
Sometimes this results in discovery errors for some services like zanata that have a hard coded low timelimit.
The database load is also somewhat high, showing ipsilon is doing something, but not clear what.
Seems like it happens quite frequent now. Zanata is experiencing fedora login issue 2-3 hours after fedora open id service restarted. Need to resolve this as soon as possible as it is affecting Zanata production system.
This has now been fixed in production, and upstream patch is up for review.
This was caused by the databases lacking any indexes, so searching the correct entries amongst more then 2 billion entries was slow.
Indexes have been created in prod, and upstream patch is at: https://pagure.io/ipsilon/pull-request/16
Thanks for reporting this issue!
This started happening again, now looking if my indices were correct, and to finish the maintenance tool to automatically clear up expired transactions/sessions, and to make a heartbeat page that prevents proxy checks from resulting in new transactions/sessions.
I think we can close this now?
New version is rolled out and we have fixed another issue that would have been causing slowness.
Login to comment on this ticket.