kentonv changed the topic of #sandstorm to: Welcome to #sandstorm: home of all things sandstorm.io. Say hi! | Have a question but no one is here? Try asking in the discussion group: https://groups.google.com/group/sandstorm-dev
<TimMc> I suspect that in most filesystems, deleting a directory is pretty permanent. A directory is nothing but file metadata. The file *contents* might still be around, though.
<isd> I suppose it depends on whether the inode has been reused. Not sure how quickly those get recycled.
<isd> Very tangentially related: haskell-capnp now supports generics, which I found myself wanting when working on ocap-merkledag (which forms the backbone of the backup system that will materialize eventually)
<isd> (though I haven't kicked out a release yet, but that will be soon)
_whitelogger has joined #sandstorm
larjona has quit [Ping timeout: 256 seconds]
larjona has joined #sandstorm
JacobWeisz[m] has quit [Quit: Bridge terminating on SIGTERM]
abliss has quit [Quit: Bridge terminating on SIGTERM]
jryans has quit [Quit: Bridge terminating on SIGTERM]
isd has quit [Quit: Bridge terminating on SIGTERM]
ill_logic has quit [Quit: Bridge terminating on SIGTERM]
Mitar has quit [Ping timeout: 245 seconds]
Mitar has joined #sandstorm
abliss has joined #sandstorm
JacobWeisz[m] has joined #sandstorm
jryans has joined #sandstorm
isd has joined #sandstorm
sam_w has left #sandstorm [#sandstorm]
<CcxWrk> TimMc: If by most you mean "FAT", in anything more modern it will be slightly more complicated.
larjona has quit [Ping timeout: 272 seconds]
larjona has joined #sandstorm
hannes_ has quit [Ping timeout: 256 seconds]
<TimMc> CcxWrk: Ext2, then -- do directory inodes just get marked as reclaimable?
crab has quit [Remote host closed the connection]
crab has joined #sandstorm
<CcxWrk> Are we talking strictly ext2 or also the (metadata) journalled variants? But AFAIK yes, directory contents is just another type of data. That's why you can end up with whole directory structures in lost+found.
<CcxWrk> This is what the "file contents" of a directory looks on ext2: http://www.science.smith.edu/~nhowe/262/oldlabs/ext2.html#direntry
<TimMc> Nice.
<TimMc> But I've never encountered a tool that offers to look for that kind of thing. :-/
<CcxWrk> I have to admit I'm not very well versed in ext* structure, it has too many corner cases I've ran into and mostly used xfs, f2fs and bit of btrfs and zfs over the years.
<TimMc> How has btrfs treated you?
<CcxWrk> Some of the analysis tools like hachoir might. But the hard thing would be, in absence of journal, actually locating valid directory entry data blocks without getting swamped by false positives.
<CcxWrk> I used btrfs on "temporary data, can download again" raid-0 setup. Didn't have a dataloss bug since linux 3.0 until HDD firmware bug caused corruption relatively recently.
<CcxWrk> It wasn't very fast. Won't probably bother with it again.
<CcxWrk> I'm looking forward to bcachefs, but it's not featureful enough yet for anything I'd like it for.
<CcxWrk> (And tbh I'm leaning towards mostly ditching Linux for FreeBSD anyway)
<TimMc> I'd been thinking of switching from zfs to btrfs so I could have easier snapshotting for backups, and do it on raspberry pi as well... but btrfs seems a little scary. :-P
<CcxWrk> I probably wouldn't bother on lower power devices. At least when you remove LVM snapshot the COW overhead goes away.
<TimMc> Oh hmm, LVM.... is that what I should be using?
<CcxWrk> If you just want to make snapshot and offload it somewhere else without having local archive? Most likely.
<TimMc> My current backup script takes a ZFS snapshot, mounts it, runs tarsnap, and unmounts and releases it. But the release fails intermittently, so I'm looking for something more reliable.
<TimMc> LVM is kind of irritating, but I could work with that.
hannes_ has joined #sandstorm
<JacobWeisz[m]> Ian
<JacobWeisz[m]> Ian, I think you broke stuff
<isd> ocdtrekkie: ?
<isd> Ah, saw the notification, will coordinate there.
<isd> ocdtrekkie: do you have a grain you can give me a backup of, for expediency of testing?
<isd> I could concoct something, but it would make things a bit easier
<isd> I think I've found the problem
<JacobWeisz[m]> Uh, yeah, I'll email it.
<isd> Thanks
<JacobWeisz[m]> Sent. Presumably for the Docs one you can reconstruct that yourself, it's using the GitWeb app. I don't have access to that actual grain, but the error was different.
<isd> Gah, looks like there are two problems
<isd> kentonv: probably makes sense to push out a hotfix ^
<isd> I am annoyed with myself for apparently not carefully testing that last minute tweak.
<kentonv> rolled back, building release
<kentonv> new release is up
<kentonv> fwiw, docs.sandstorm.io was broken
<isd> Yeah, aware.
<isd> Can you get me a backup of that grain? I'm pretty sure the patch above would fix the problem, but I'd like to test.
<kentonv> ironically, it is not my grain, so I can't download a backup. I think it might be Asheesh's.
xet7 has joined #sandstorm
asheesh has joined #sandstorm
<isd> Apparently the bad update never hit my server.
<asheesh> Hey isd & JacobWeisz[m] -- I saw https://github.com/sandstorm-io/sandstorm/issues/3472 and just wanted to see if it's still useful for me to send a backup of the docs.sandstorm.io grain?
<asheesh> Looks like docs are doing OK, so it's not urgent; that's good.
<isd> Yeah, we've already rolled back the regression -- per my comment above we got to it fast enough that my own box never saw the bad update
<asheesh> :)
<isd> But I did end up deploying the docs on my own box, so I think no need for a backup from you
<asheesh> isd: Cool! I did make the backup, so here's a Davros link for you if you like :) https://alpha.sandstorm.io/shared/t1G2BHJHLKv-bNAGiNKlaMGgJSFtfZ9iCyQf3mhbwR5
<isd> Thanks!
<asheesh> Ciao for now! Can chat more by email later.
asheesh has left #sandstorm [#sandstorm]
<kentonv> um did asheesh just share a backup in a public channel. :/
<simpson> It's okay, I covered my eyes.
<isd> Hah
<isd> I don't think there's actually any private info in there though? it's just a git repo with the static docs in it.
<kentonv> yeah I guess that's true...
<kentonv> FWIW I think the release was up for 23-ish hours which means most people probably did get it.
<kentonv> (since people's servers check for updates every 24 hours)
<isd> That was my suspicion
<isd> Anyway, the fix does appear to work as expected.
<JacobWeisz[m]> Should we test any other web publishing apps in case others have other ways of doing things?
<isd> Probably. Do we have a list of apps that use static publishing somewhere?
<JacobWeisz[m]> Doubtful, we recommend that good Sandstorm apps should though.
<JacobWeisz[m]> Probably the big ones to verify are WordPress and Ghost?
xet7 has quit [Remote host closed the connection]
<isd> Noted those on the issue.