ceph - cephfs - 2024-11-21

Timestamp (UTC)Message
2024-11-21T10:22:30.532Z
<Igor Golikov> Hi folks, i need some help with snapshots. I want to create snapshot, so I enabled it ( I see `flags	12 joinable allow_snaps allow_multimds_snap` in the `ceph fs get a` output). but when I run `ceph fs snapshot create...` it says Invalid command. I have the `.snap` folder in my fs. What do I miss?
2024-11-21T10:23:24.092Z
<Venky Shankar> Can you share the full command you ran  @Igor Golikov?
2024-11-21T10:23:45.360Z
<Venky Shankar> `.snap` is virtual do its always iteratable even without snapshots.
2024-11-21T10:27:11.571Z
<Skylar Kelty> Hi all,

I'm following the DR procedure at <https://docs.ceph.com/en/reef/cephfs/disaster-recovery-experts/> to recover a ~2PiB FS.
The problem is, scan_extents doesn't _*seem*_ to be doing anything.
strace is just showing a constant stream of:
```
futex(0x556dd24fa618, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7ffcaa19e278, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, FUTEX_BITSET_MATCH_ANY) = 0
futex(0x7ffcaa19e220, FUTEX_WAKE_PRIVATE, 1) = 0
write(10, "c", 1)                       = 1
```
No data is being written to the new metadata pool.

Question: do all "scan_extents" workers need to be running on the same node or can you spread them out across multiple nodes as I have?
2024-11-21T10:28:01.920Z
<Skylar Kelty> I didn't think "snapshot create" was a command, don't you just "mkdir .snap/<snap name>" to create them?
2024-11-21T10:28:34.167Z
<Skylar Kelty> Hi all,

I'm following the DR procedure at <https://docs.ceph.com/en/reef/cephfs/disaster-recovery-experts/> to recover a ~2PiB FS.
The problem is, scan_extents doesn't _*seem*_ to be doing anything.
strace is just showing a constant stream of:
```futex(0x556dd24fa618, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7ffcaa19e278, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, FUTEX_BITSET_MATCH_ANY) = 0
futex(0x7ffcaa19e220, FUTEX_WAKE_PRIVATE, 1) = 0
write(10, "c", 1)                       = 1```
No data is being written to the new metadata pool.

Question: do all "scan_extents" workers need to be running on the same node or can you spread them out across multiple nodes as I have?
2024-11-21T10:28:55.158Z
<Igor Golikov> I tried the latter, but it says "readonly folder"
2024-11-21T10:29:00.618Z
<Venky Shankar> oh, that's an interface to take snapshots on cephfs subvolumes.
2024-11-21T10:29:12.679Z
<Igor Golikov> cannot create regular file '.snap/2': Read-only file system
2024-11-21T10:29:28.627Z
<Igor Golikov> this is what I get when i try to mkdir under .snap
2024-11-21T10:32:12.331Z
<Igor Golikov> OH i get it
2024-11-21T10:32:28.535Z
<Igor Golikov> i was trying to mkdir when I am in the .snap folder
2024-11-21T10:33:52.284Z
<Venky Shankar> ah kk
2024-11-21T10:34:09.814Z
<Venky Shankar> what was the issue with `fs snapshots create` command?
2024-11-21T10:34:21.131Z
<Venky Shankar> what was the issue with `fs snapshot create` command?
2024-11-21T10:34:24.519Z
<Igor Golikov> jsut got "unknown command"
2024-11-21T10:35:54.681Z
<Igor Golikov> but now I can't delete .snap...
rm -rf .snap/ gives me "rm: cannot remove '.snap/2/2': Read-only file system"
2024-11-21T10:36:05.400Z
<Igor Golikov> do we have any docs on how to use snaps?
2024-11-21T10:36:07.588Z
<Venky Shankar> ```./bin/ceph fs subvolume create a sub0
➜  build git:(main) ✗ ./bin/ceph fs subvolume snapshot create a sub0 snap0
➜  build git:(main) ✗ ./bin/ceph fs subvolume snapshot ls a sub0
[
    {
        "name": "snap0"
    }
]```
2024-11-21T10:36:38.993Z
<Venky Shankar> > do we have any docs on how to use snaps?
There is a dev guide. @gregsfortytwo recently created a tracker to have admin docs for snapshots (which was lacking!).
2024-11-21T10:36:52.413Z
<Skylar Kelty> I think you target the snap "rm .snap/<snap name>"
2024-11-21T10:37:18.755Z
<Venky Shankar> @Igor Golikov <https://tracker.ceph.com/issues/68974> (nice timing btw 🙂)
2024-11-21T10:38:47.545Z
<Igor Golikov> @Skylar Kelty I have now .snap/2/2 (the last 2 is a file) and I cant delete neither .snap nor .snap/2
2024-11-21T10:39:12.034Z
<Igor Golikov> is aways says `rm: cannot remove '.snap/2/2': Read-only file system`
2024-11-21T10:39:45.231Z
<Skylar Kelty> try rmdir instead of rm
2024-11-21T10:40:11.565Z
<Igor Golikov> oh... nice
2024-11-21T10:40:13.080Z
<Igor Golikov> thanks!!
2024-11-21T10:40:31.448Z
<Venky Shankar> @Igor Golikov `rmdir .snap/2`
2024-11-21T10:41:12.519Z
<Igor Golikov> so if i want to create a snapshot now i can just copy stuff under .snap/<snapname> ?
2024-11-21T10:41:41.724Z
<Venky Shankar> there is no need to copy
2024-11-21T10:42:13.781Z
<Venky Shankar> once you take a snap under a dir, everything under that dir (snapped) is accessible under the snapshot dir (under .snap/<snapname>).
2024-11-21T10:42:21.379Z
<Igor Golikov> ohhh
2024-11-21T10:42:56.757Z
<Igor Golikov> got it, so running the `mkdir .snap/<snapname>` itself creates the snapshot
2024-11-21T10:43:03.309Z
<Venky Shankar> yeh
2024-11-21T10:43:06.540Z
<Igor Golikov> got it.
2024-11-21T10:43:08.580Z
<Igor Golikov> 10
2024-11-21T10:43:11.872Z
<Igor Golikov> 10x
2024-11-21T12:16:46.943Z
<Skylar Kelty> Hi all,

I'm following the DR procedure at <https://docs.ceph.com/en/reef/cephfs/disaster-recovery-experts/> to recover a ~2PiB FS.
The problem is, scan_extents doesn't _*seem*_ to be doing anything.
Question: ~~do all "scan_extents" workers need to be running on the same node or can you spread them out across multiple nodes as I have?~~
Read the [source code](https://github.com/ceph/ceph/blob/main/src/tools/cephfs/DataScan.cc#L655) and checked - they can run on multiple machines
2024-11-21T12:16:48.607Z
<Skylar Kelty> Hi all,

I'm following the DR procedure at <https://docs.ceph.com/en/reef/cephfs/disaster-recovery-experts/> to recover a ~2PiB FS.
The problem is, scan_extents doesn't _*seem*_ to be doing anything.
Question: ~~do all "scan_extents" workers need to be running on the same node or can you spread them out across multiple nodes as I have?~~
Read the [source code](https://github.com/ceph/ceph/blob/main/src/tools/cephfs/DataScan.cc#L655) and checked - they can run on multiple machines
2024-11-21T12:16:52.223Z
<Skylar Kelty> Hi all,

I'm following the DR procedure at <https://docs.ceph.com/en/reef/cephfs/disaster-recovery-experts/> to recover a ~2PiB FS.
The problem is, scan_extents doesn't _*seem*_ to be doing anything.
Question: ~~do all "scan_extents" workers need to be running on the same node or can you spread them out across multiple nodes as I have?~~
Read the [source code](https://github.com/ceph/ceph/blob/main/src/tools/cephfs/DataScan.cc#L655) and checked - they can run on multiple machines
2024-11-21T12:40:41.291Z
<Venky Shankar> `sca_extents` is going to loop objects in tile file system data pool, so, the time taken is going to rely on that many number of objects.
2024-11-21T12:41:02.313Z
<Venky Shankar> and yes, you can run multiple workers to speed it up 👍
2024-11-21T12:43:51.147Z
<Venky Shankar> `scan_extents` is going to loop objects in tile file system data pool, so, the time taken is going to rely on that many number of objects.
2024-11-21T13:32:54.005Z
<Rishabh Dave> jitsi is too slow for me. i can't join it since last few minutes. anyone else facing same issues?
2024-11-21T13:33:57.501Z
<jcollin> no, it's good here.
2024-11-21T13:34:47.424Z
<Rishabh Dave> i can't join, it's stuck in loading mode. ping to jitsi are take >200 ms and to google it takes only few ms.
2024-11-21T13:37:47.561Z
<Rishabh Dave> finally, was able to join now.
2024-11-21T14:54:32.399Z
<gregsfortytwo> rmdir .snap/2
2024-11-21T14:54:47.815Z
<Igor Golikov> Hi folks, i need some help with snapshots. I want to create snapshot, so I enabled it ( I see `flags	12 joinable allow_snaps allow_multimds_snap` in the `ceph fs get a` output). but when I run `ceph fs snapshot create...` it says Invalid command. I have the `.snap` folder in my fs. What do I miss?
2024-11-21T16:01:07.404Z
<Hutch> Hi All, im trying to use the cephfs root_squash option on mounting cephfs but it does not seem to be working. the client system can still remove and delete files as the root user
2024-11-21T16:06:45.321Z
<Hutch> the cluster is running Quincy 17.2.7 and this is my keyring, tried both kernel and Fuse mounts

```client.home
        key: XXXXXXXX
        caps: [mds] allow rw fsname=cephfs root_squash path=/home
        caps: [mgr] allow r
        caps: [mon] allow r
        caps: [osd] allow *```
2024-11-21T16:07:45.406Z
<Hutch> i also tried this keyring

```[client.home]
        key = AXXXXX
        caps mds = "allow rw fsname=cephfs root_squash, allow rw fsname=cephfs path=/home"
        caps mgr = "allow r"
        caps mon = "allow r"
        caps osd = "allow *" ```
2024-11-21T16:09:06.051Z
<Hutch> i also tried this keyring

```client.home
        key: XXXXXXXX
        caps: [mds] allow rw fsname=cephfs root_squash, allow rw fsname=cephfs path=/home
        caps: [mgr] allow r
        caps: [mon] allow r
        caps: [osd] allow *```

Any issue? please create an issue here and use the infra label.