Skip to content
This repository has been archived by the owner on Jul 7, 2020. It is now read-only.

Pod devices for topology get stuck, if pods are restarted. #642

Open
andresolotowas opened this issue Feb 17, 2020 · 0 comments
Open

Pod devices for topology get stuck, if pods are restarted. #642

andresolotowas opened this issue Feb 17, 2020 · 0 comments

Comments

@andresolotowas
Copy link

andresolotowas commented Feb 17, 2020

Hey guys, faced the following issue, would appreciate your help!

A few days ago I was doing a GlusterFS deploy into my cluster, so following the instructions Ive created loop devices like this

mkdir home/gluster; dd if=/dev/zero of=/home/gluster/image bs=1M count=10000; losetup /dev/loop0 /home/gluster/image

Given this I was able to load the topology later with /dev/loop0 devices successfully. Ive removed Heketi deploy, Gluster pods and whatever else created and started again from scratch.
Now I try to repeat all the way done before, I what I now see is - on new Gluster pods, there are no longer "/home/gluster/image" files. This is OK, since that pods were removed, but it was still left on the nodes, as I understand. So now, in the new Gluster pods I still see the "/dev/loop0" that is explained as

[root@p-4v84 /]# losetup -a
/dev/loop0: [0203]:2235485 (/home/gluster/image)

so referring a missing file, actually, from that old pod I've removed, and what I see else is:
[root@p-4v84 /]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
vdb 254:16 0 458K 1 disk
loop0 7:0 0 9.8G 0 loop
├─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc_tdata 253:1 0 2G 0 lvm
│ └─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc-tpool 253:2 0 2G 0 lvm
│ ├─vg_f581b805ccd1f865b901e290ebaba0ff-brick_edd1f9ff0dc51aa426c6d0e399c011bc 253:4 0 2G 0 lvm /var/lib/heketi/mounts/vg_f581b805ccd1f865b901e290ebaba0ff/brick_edd1f9ff0dc51aa426c6d0e399c011bc
│ └─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc 253:3 0 2G 0 lvm
└─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc_tmeta 253:0 0 12M 0 lvm
└─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc-tpool 253:2 0 2G 0 lvm
├─vg_f581b805ccd1f865b901e290ebaba0ff-brick_edd1f9ff0dc51aa426c6d0e399c011bc 253:4 0 2G 0 lvm /var/lib/heketi/mounts/vg_f581b805ccd1f865b901e290ebaba0ff/brick_edd1f9ff0dc51aa426c6d0e399c011bc
└─vg_f581b805ccd1f865b901e290ebaba0ff-tp_edd1f9ff0dc51aa426c6d0e399c011bc 253:3 0 2G 0 lvm
vda 254:0 0 50G 0 disk
└─vda1 254:1 0 50G 0 part /var/lib/glusterd

As I try to understand - this is a mount from the node itself, or something like this. And the issue is - I cannot either use this /dev/loop0 device again or I cannot umount it, losetup -d, or rm, or wipefs -a or whatever else I've already tried. No results.
This is what I get from Heketi

[root@deploy-heketi-6c687b4b84-llrg9 /]# heketi-cli device add --name=/dev/vda/loop0 --node=3d7bea391ff3e6565f70c8053e8de8b3\
Error: Initializing device /dev/vda/loop0 failed (already initialized or contains data?): WARNING: Failed to connect to lvmetad. Falling back to device scanning.
Device /dev/vda/loop0 not found.

Mostly as same as I get locally on the pod

[root@p-4v84 /]# pvscan
WARNING: Failed to connect to lvmetad. Falling back to device scanning.
No matching physical volumes found

And here is what topology load feeds me back:

[root@deploy-heketi-6c687b4b84-llrg9 tmp]# heketi-cli topology load --json=topology.json
Found node p-4v84 on cluster b455b24ff6598025c7a04f8b1ec8e4cc
Adding device /dev/loop0 ... Unable to add device: Initializing device /dev/loop0 failed (already initialized or contains data?): WARNING: Failed to connect to lvmetad. Falling back to device scanning.
Can't open /dev/loop0 exclusively. Mounted filesystem?
Can't open /dev/loop0 exclusively. Mounted filesystem?
Found node p-4v8h on cluster b455b24ff6598025c7a04f8b1ec8e4cc
Adding device /dev/loop0 ... Unable to add device: Initializing device /dev/loop0 failed (already initialized or contains data?): WARNING: Failed to connect to lvmetad. Falling back to device scanning.
Can't open /dev/loop0 exclusively. Mounted filesystem?
Can't open /dev/loop0 exclusively. Mounted filesystem?
Found node p-4v8i on cluster b455b24ff6598025c7a04f8b1ec8e4cc
Adding device /dev/loop0 ... Unable to add device: Initializing device /dev/loop0 failed (already initialized or contains data?): WARNING: Failed to connect to lvmetad. Falling back to device scanning.
Can't open /dev/loop0 exclusively. Mounted filesystem?
Can't open /dev/loop0 exclusively. Mounted filesystem?

Do you have any clues how can I maybe reuse this device? Or create a new one maybe (but how to be with a disk space this case if I leave this still alive)?

Would be really glad and appreciate any ideas around this.

PS. And one more thing unclear - would be nice to get it clarified - why does the GlusterFS pod come without a device ready to go into topology? I dont see it from the box, I dont see it in "gk-deploy" script. Why do I have to create it by hands?

Thank you in advance!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant