Recent searches
Search options
One problem with the 4.3 docker-based #Mastodon is a lot of 504 gateway timeout errors popping up for no obvious reason (especially with the notification timeline that can take ages to fill in). Seems to be a common reverse proxy problem but it can be quite annoying. This seems to start happening if the system is slowed down for some reason -- like taking backups of the voluminous /public/system directories that are just loaded with hundreds of thousands of tiny files. Even after the backup finishes, #Mastodon doesn't seem to recover properly and the 504s can continue. Still looking into this.
Are the thousands-of-tiny-files split into many small subdirs, or just individual directories with thousands of files in them?
If the latter, then depending on the #filesystem type, Linux can be quite slow at looking up and opening files in those dirs - if the proxy (or whatever) is expecting near-instantaneous results, it might be sad.
ext2/3/4 are problematic with the tons-of-files-in-a-dir issue, for example.
@cazabon Several layers of directories, with the payload in the bottom layers being up to 1000 little files per directory.
Okay, 1000 files is around where it makes sense to start thinking of splitting directories up with the ext? filesystems if you're going to keep adding files to those dirs. If they don't go much over 1000 each, it shouldn't be a major issue.
@cazabon Thing is, I didn't have this problem with 4.2. But right now I've had to move to Firefox to use 4.3 without a lot of hassle. I don't understand what's going on.
@cazabon Chrome Beta was working fine until I ran a long backup from that system, then afterward it is all messed up again -- even after restarts and reboots. But Firefox is working fine with it. Very bizarre.
Truly bizarre. I have difficulty seeing Chrome start erroring out on valid sites/connections.
This is Chrome (or FF) running on a separate machine from the one hosting the Docker containers? You're not testing the containers on your main workstation/personal system?
@cazabon Of course. The browsers are on a desktop, the servers are on ... the servers.
I figured, but had to ask.
Are any of the processes in the containers dying? Or from the logs etc, it just looks like those processes are hunkey-dorey?
@cazabon The containers themselves seem to mostly be OK. It's possible for them to get confused and have duplicates started up if you're not careful, especially if the system is booted either cleanly (or worse) not cleanly. I haven't worked out the best procedures for every contingency yet.
@cazabon Chrome doesn't throw the errors. But accessing the server via Chrome -- when this starts to happen -- you start seeing 504 (gateway timeout) errors -- which would be the reverse proxy (nginx).
Aha - thanks for the clarification/correction. I'll shut up now.
@cazabon Because sometimes it starts working again with Chrome for no apparent
reason, I do suspect something to do with caching.