On lemmy.today, I don’t appear to be seeing posts created in the past two days as of this writing on either !NonCredibleDefense@sh.itjust.works or !Games@sh.itjust.works. I am subscribed to both communities. I have not done an search to determine whether other subscribed sh.itjust.works communities are affected, though I assume that to be the case.
They are visible from lemmy.world, so the posts are propagating to at least some other federated instances.
sh.itjust.works presently appears to be running Lemmy 0.19.5 (versus lemmy.today’s 0.19.4 and lemmy.world’s 0.19.3). However, I do not think that this is a version-related problem, at least not alone, as !Ukraine@sopuli.xyz is showing up fine on lemmy.today.
https://lemmy.today/instances lists sh.itjust.works as a federated instance as of this writing.
https://sh.itjust.works/instances lists lemmy.today as a federated instance as of this writing.
Lemmy.today has also been responding very slowly to me over the past hour or so, and was frequently showing connection timeouts and gateway error pages when trying to load pages. Other instances appear to be working normally. That may be entirely unrelated, but I thought that I’d mention it, as it’s unusual and at least might be related.
Note, on the second, probably-unrelated “slow lemmy.today over the past hour” thing, looking at the behavior I see with
tcpdump, I see periods where there’s no response to SYN packets at all in a quick test. So I’d assume that whatever is going on there is either related to network connectivity (I don’t know whether you have a reverse proxy of some sort set up, but might be between that and your actual host) or if not on the network, at the kernel level on lemmy.today. My machine was sending a SYN on connection initiation, but getting no SYN+ACK packet back. My understanding is that SYN+ACK should be sent by the lemmy.today kernel prior to the userspace-level involvement in the connection process; the lemmy.today kernel should be doing that process independent of any issues in userspace, shouldn’t need anaccept()call.It’s responding normally at the moment. And I’m not super-worried about the latter issue, as far all I know it’s some sort of transient problem with Hetzer’s infrastructure; just wanted to mention it, as it might be relevant to the prior problem.