broken import? about pacman.store

[RubenKelevra]

All issues should be fixed. And the cluster is running as it should (2 servers are catching up with metadata).

I will replace the import server with one with SSDs, since the issue is now better but still way below my expectations. I also contacted the support, since this should be an imminent hardware failure.

[RubenKelevra]

No it's not a local issue. I have seen this too, and fixed it for one package. Though it was limited to this one.

I'll replay the last 2 month week again as a diff update - this should fix it.

Edit: I decided to run the full archived diff-log to make sure there's no second import issue longer back.

[RubenKelevra]

I rolled the full diff through, took longer than expected (8 hours).

My hosting provider has had some serious issues with my disk performance on this server, so I need to run an optimizing job. Don't know how long this is going to take, but the performance is AWFUL - so it really needs to run.

Status can be seen here, as soon as the operation continues it should contain a full mirror.

[rpodgorny]

ok. will wait and report later.

anyway, this kind of reminds me it may be a good idea to think about how to avoid dependence on single server in the future (to make the ipfs's "distributed web" claim become reality). is there a way to update ipns records from multiple (backup) locations (servers)?

[RubenKelevra]

@rpodgorny yes and no. You can use the same key, but currently you would have to wait for the old record to expire. See ipfs/kubo#6683

And seeding IPNS with multiple nodes would also be great:
ipfs/kubo#4536

But we're in the beginning. This is still alpha software.

When the first mentioned limitation is fixed, I could run the software on two nodes which can write to the cluster, and it should eventually lead to both nodes publishing the same CID to the same IPNS, at different times.

So if the second node resolves the IPNS and it already got updated, it won't push another update. But if it's different, it will push an update.

Concurrent writing to the cluster is no issue and concurrent publishing of IPNS records (in theory) neither. :)

[rpodgorny]

confirming it works for me now. thanks!

[RubenKelevra]

Very well! :)

I'm replacing the main server in the next hours, to fix the speed issues as well. :)