Difference between revisions of "2024/11/29"
Jump to navigation
Jump to search
| (6 intermediate revisions by the same user not shown) | |||
| Line 2: | Line 2: | ||
Nobody can post or follow on TootCat, though the page seems to reload and remote updates are coming in. When trying to post, a "500" error appears in the lower right corner. | Nobody can post or follow on TootCat, though the page seems to reload and remote updates are coming in. When trying to post, a "500" error appears in the lower right corner. | ||
==Diagnosis== | ==Diagnosis== | ||
| − | * '''08:50 EST''' Tentatively, the web servers and the database are fine -- but I need to discuss this with our web engineer (Dan) to confirm my understanding and work out a fix. I don't want to go making things worse with a poorly-conceived kluge that doesn't actually work. | + | * '''10:27 EST''' ...and yet! It turns out that, by a curious coincidence, the root volume somehow got all filled up. (This is ''not'' where we keep the database or media files.) Once Dan removed a bunch of old, unneeded snapshot files, everything started working again (although Hetz had also reported that things were recovering on their end as well; hard to say if this would have worked before that point). |
| + | ** Dan is tweaking the garbage-collection routines to catch this sort of thing in the future. | ||
| + | * '''09:55 EST''' I've confirmed that the cloud consoles are no longer accessible (disabled "to help us recover"), so this is definitely their issue. | ||
| + | ** ...at least for now; if things don't start working again once they're done, then we'll have to resume investigating. | ||
| + | * '''09:38 EST''' New posts from outside are also not coming in (last new message was last night at 19:48 EST) -- which would be consistent with a blockage between the back-end server and the rest of the internet. So we can see the server's current state, but it isn't communicating with the rest of fedi. I can access most of the cpanel, including moderation, but for some reason the [https://toot.cat/admin/dashboard admin panel] also returns an error; I'm not sure how that fits with my theory. | ||
| + | * '''08:50 EST''' Tentatively, the web servers and the database are fine -- but I need to discuss this with our web engineer (Dan) to confirm my understanding and work out a fix. I don't want to go making things worse with a poorly-conceived kluge that doesn't actually work. It looks like the load-balancer/firewall is the only thing affected, but I don't have a firm enough grasp of our architecture to be sure that this is the case or to be sure of how to temporarily work around it. | ||
| + | ** '''Note''': I figured this out before they disabled the cloud consoles. | ||
* '''08:14 EST''' The problem may be a [https://status.hetzner.com/incident/c7a683c0-8216-45ef-87dc-1c8574ba714d service outage at Hetzner]. We're checking possibilities. | * '''08:14 EST''' The problem may be a [https://status.hetzner.com/incident/c7a683c0-8216-45ef-87dc-1c8574ba714d service outage at Hetzner]. We're checking possibilities. | ||
** This was originally "API failure (datacenters & locations)" but later changed to "Failures on API and Console". | ** This was originally "API failure (datacenters & locations)" but later changed to "Failures on API and Console". | ||
| + | ** Updates from Hetzner: | ||
| + | *** '''2024-11-29 15:09 UTC+0''' The recovery is in progress. The systems are recovering slowly right now. | ||
| + | *** '''2024-11-29 14:08 UTC+0''' It is currently not possible to access the Cloud Console. This is intended to help us recover. | ||
| + | *** '''2024-11-29 12:41 UTC+0''' Currently it is not possible to create or modify resources in the Hetzner Cloud. Already running resources are not affected. | ||
| + | **** ...except it looks like our firewall/load-balancer is out of commission. -W. | ||
Latest revision as of 15:44, 29 November 2024
Problem
Nobody can post or follow on TootCat, though the page seems to reload and remote updates are coming in. When trying to post, a "500" error appears in the lower right corner.
Diagnosis
- 10:27 EST ...and yet! It turns out that, by a curious coincidence, the root volume somehow got all filled up. (This is not where we keep the database or media files.) Once Dan removed a bunch of old, unneeded snapshot files, everything started working again (although Hetz had also reported that things were recovering on their end as well; hard to say if this would have worked before that point).
- Dan is tweaking the garbage-collection routines to catch this sort of thing in the future.
- 09:55 EST I've confirmed that the cloud consoles are no longer accessible (disabled "to help us recover"), so this is definitely their issue.
- ...at least for now; if things don't start working again once they're done, then we'll have to resume investigating.
- 09:38 EST New posts from outside are also not coming in (last new message was last night at 19:48 EST) -- which would be consistent with a blockage between the back-end server and the rest of the internet. So we can see the server's current state, but it isn't communicating with the rest of fedi. I can access most of the cpanel, including moderation, but for some reason the admin panel also returns an error; I'm not sure how that fits with my theory.
- 08:50 EST Tentatively, the web servers and the database are fine -- but I need to discuss this with our web engineer (Dan) to confirm my understanding and work out a fix. I don't want to go making things worse with a poorly-conceived kluge that doesn't actually work. It looks like the load-balancer/firewall is the only thing affected, but I don't have a firm enough grasp of our architecture to be sure that this is the case or to be sure of how to temporarily work around it.
- Note: I figured this out before they disabled the cloud consoles.
- 08:14 EST The problem may be a service outage at Hetzner. We're checking possibilities.
- This was originally "API failure (datacenters & locations)" but later changed to "Failures on API and Console".
- Updates from Hetzner:
- 2024-11-29 15:09 UTC+0 The recovery is in progress. The systems are recovering slowly right now.
- 2024-11-29 14:08 UTC+0 It is currently not possible to access the Cloud Console. This is intended to help us recover.
- 2024-11-29 12:41 UTC+0 Currently it is not possible to create or modify resources in the Hetzner Cloud. Already running resources are not affected.
- ...except it looks like our firewall/load-balancer is out of commission. -W.