-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Stuck in "empty" #matrix-dev:matrix.org room #1933
Comments
We've been seeing a few reports of scenarios like this over the last few
weeks - it sounds like there may be a regression in 0.19.0.
Is there any chance you can share the full homeserver logs with me via
Matrix or similar so we can investigate what happened?
@richvdh this may be relevant to your current interests...
|
Sure thing, I've reached out via matrix to you directly. |
we don't have logs due to a config error, but hopefully this is going to
turn out to be the same as the "suddenly kicked out of Matrix HQ"
problem that boli.rocks was encountering earlier today...
|
user error on my end, looks like the Arch packager is responsible and was using journalctl. |
logs have been extracted from the depths of systemd - @richvdh i have them on my laptop if you want to peruse tomorrow. |
tbp this is #1935 |
Hum; having looked at storrgie's logs, and thought about this a little more, I'm not entirely sure it is #1935. In this case the problem seems to have happened between 2017-02-04 15:42:13, when storrgie joined #matrix-dev and immediately left again (as seen on matrix.org), and Feb 18 06:40:03 when his client sent another join request on the room, which is rejected. At some point in that period storrgie's HS has gone from thinking he was not a member of the room, to (at least in part) thinking he was. @storrgie: would you mind running a query for me to help diagnose:
|
Investigation of storrgie's db shows that his homeserver has recorded a very limited amount of state for matrix-dev, and in particular doesn't include his leave event:
|
So it looks like storrgie's db had got itself into a particular mess, which I don't think will be a common case. I think what happened is:
It remains a bit unclear what happened on 11 Feb to end up in such an inconsistent state. The solution in this case was:
I have opened #1940 to track the more general reports of people being spuriously rejoined to rooms. |
Very likely my fault due to this being my evaluation/deployment order of events:
I did not migrate the sqlite to postgres, I just deleted and started over. This was stupid of me. I know that before (on sqlite), and after (on postgres) I was using the same MXID. This likely compounded the issue. |
I would not be offended if you closed this @richvdh, it's mainly up for people to be informed when they do silly things during their initial deployment. |
thanks. There are still some outstanding questions about why the server behaved the way it did when you rejoined. In particular, why it accepted your join event but apparently not your leave event (or anything thereafter). I'll keep it open for now so that if we get something similar in future we might be able to get some more info. |
I had the same problem, but I had no entry in
After that I could rejoin matrix-dev |
@rubo77 sounds like a different problem |
Same problem here: #3001 |
This bug was meant to be quite specific to storrgie's problems. I'm not sure that keeping it open any longer has any value, since we're unlikely to go back and investigate that particular sequence of events. A lot of the underlying problems would be solved by fixing #3422. |
I've federated my homeserver about two weeks ago and one of my initial tests was to join
#matrix-dev:matrix.org
, which worked fine. Somewhere along the line the room turned into "Empty Room" on my room list, with an identifier of!XqBunHwQIXUiqCaoxq:matrix.org
.Now it seems I am unable to join
#matrix-dev:matrix.org
, and I cannot leave/forget the "Empty Room", the error message beingM_UNKNOWN: Room '!XqBunHwQIXUiqCaoxq:matrix.org' does not exist
.this may be user error, I am just not sure of where to report it.
The text was updated successfully, but these errors were encountered: