2024-11-14T12:06:33.528Z | <Venky Shankar> @rzarzynski @Laura Flores Is there a fix for the `cluster [ERR] Failed to load ceph-mgr modules: prometheus" in cluster log` failure? e.g.: <https://pulpito.ceph.com/teuthology-2024-11-01_20:24:44-fs-main-distro-default-smithi/7976482> |
2024-11-14T12:23:41.362Z | <rzarzynski> @Laura Flores: I guess this is just another instance of the generic lack of sync between loading module & processing commands. Am I correct? |
2024-11-14T13:19:59.748Z | <Markuze> Hey falks, I m trying to mount two different file system on a VM, Each mount separatly works bat together `ls` and `df` hang.
```[global]
fsid = 7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
mon_host = [v2:192.168.122.251:40471/0,v1:192.168.122.251:40472/0] [v2:192.168.122.251:40473/0,v1:192.168.122.251:40474/0] [v2:192.168.122.251:40475/0,v1:192.168.122.251:40476/0]
This command.
FSID=7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
SECRET=AQA68jVnTZTnLRAABiwd51tQqn4qt/n9Ka9wxA==
USER=dev-03
PORT_A=40472
PORT_B=40474
sudo mount -t ceph $USER@$FSID.a=/ /mnt/test -o mon_addr=192.168.122.251:$PORT_A,secret=$SECRET
sudo mount -t ceph $USER@$FSID.b=/ /mnt/scratch -o mon_addr=192.168.122.251:$PORT_B,secret=$SECRET``` |
2024-11-14T13:20:16.639Z | <Markuze> Hey falks, I m trying to mount two different file system on a VM, Each mount separatly works but together `ls` and `df` hang.
```[global]
fsid = 7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
mon_host = [v2:192.168.122.251:40471/0,v1:192.168.122.251:40472/0] [v2:192.168.122.251:40473/0,v1:192.168.122.251:40474/0] [v2:192.168.122.251:40475/0,v1:192.168.122.251:40476/0]
This command.
FSID=7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
SECRET=AQA68jVnTZTnLRAABiwd51tQqn4qt/n9Ka9wxA==
USER=dev-03
PORT_A=40472
PORT_B=40474
sudo mount -t ceph $USER@$FSID.a=/ /mnt/test -o mon_addr=192.168.122.251:$PORT_A,secret=$SECRET
sudo mount -t ceph $USER@$FSID.b=/ /mnt/scratch -o mon_addr=192.168.122.251:$PORT_B,secret=$SECRET``` |
2024-11-14T13:20:45.428Z | <Markuze> Any ideas why? |
2024-11-14T13:47:35.883Z | <Markuze> Hey folks, I m trying to mount two different file system on a VM, Each mount separately works but together `ls` and `df` hang.
```[global]
fsid = 7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
mon_host = [v2:192.168.122.251:40471/0,v1:192.168.122.251:40472/0] [v2:192.168.122.251:40473/0,v1:192.168.122.251:40474/0] [v2:192.168.122.251:40475/0,v1:192.168.122.251:40476/0]
This command.
FSID=7a8bc9a1-b83f-46f9-8eeb-d7bd50012563
SECRET=AQA68jVnTZTnLRAABiwd51tQqn4qt/n9Ka9wxA==
USER=dev-03
PORT_A=40472
PORT_B=40474
sudo mount -t ceph $USER@$FSID.a=/ /mnt/test -o mon_addr=192.168.122.251:$PORT_A,secret=$SECRET
sudo mount -t ceph $USER@$FSID.b=/ /mnt/scratch -o mon_addr=192.168.122.251:$PORT_B,secret=$SECRET``` |
2024-11-14T14:38:21.217Z | <Markuze> cc: @Venky Shankar |
2024-11-14T15:06:37.451Z | <Chenhao Xv> Hello, I've posted at irc but it seems that here is the right place to seek for help. I'm adding new features to Ceph and want to test its performance on real machines. I wonder how to deploy a Ceph cluster from source code. I've successfully built code into deb packages but cephadm still requires a custom Ceph container to bootstrap. However, I fail to find any useful documentation concerning building images from source. What's the most common way developers deploy from source code without getting access to sepia? |
2024-11-14T15:12:57.466Z | <John Mulligan> <https://github.com/ceph/ceph/tree/main/container>
This in-tree method of building a container image is relatively new. But should become the main method in the near future. |
2024-11-14T15:14:02.300Z | <John Mulligan> Do note that the images the ceph upstream builds is based on rpms and centos. However, the principle should be the same if you want to make your own image based on debs |
2024-11-14T15:15:39.334Z | <John Mulligan> ... it's just a bunch more work 🙂 |
2024-11-14T15:22:47.002Z | <Chenhao Xv> Thanks a lot! I'll try it myself later. |
2024-11-14T15:40:40.272Z | <Md Mahamudur Rahaman Sajib> I am posting my question here as well, to get attention from developers.
<https://ceph-storage.slack.com/archives/C04LVQMHM9B/p1731598053032089> |
2024-11-14T16:36:06.255Z | <Venky Shankar> anything in kernel ring buffer @Markuze? |
2024-11-14T16:42:10.428Z | <Venky Shankar> I'm sure mounting two file different file systems in a single node works - I've tried that before. So, not sure what'd going on. Try checking `dmesg` . |
2024-11-14T17:03:08.364Z | <Markuze> I just needed to use `v2`. V1 gets confused on the client side for some reason.
Jeff was [using](<https://jtlayton.wordpress.com/2021/11/29/testing-the-linux-kernel-cephfs-client-with-xfstests/>) two IP to get around it
Anyway this works now, but only with `V2` , it seems that IP/PORT of the monitor are communicated in the V1 protocol? I tried proxiying and NATing the second mount address first, but it still knew the actual monitor IP. |
2024-11-14T17:49:31.273Z | <Bailey Allison> not sure if this is the correct channel but both the debian-octopus and rpm-octopus repositories are missing from [download.ceph.com](http://download.ceph.com), is there any maintenance being done on the website or anything to cause this ? |
2024-11-14T17:52:43.308Z | <Casey Bodley> quoting Dan Mick from <#C1HFJ4VTN|>:
> I just moved them because the 3TB disk was out of space. Ffs who is still relying on octopus? I can symlink them back but we need to.have a conversation about how much we want to save....there's still pre-argonaut stuff there |
2024-11-14T17:55:26.887Z | <Bailey Allison> hahahaha true, sadly we still have a couple people on it.....thanks for the update ! was curious where it was only octopus that was missing |