Bug 1722390 - "All subvolumes are down" when all bricks are online
Summary: "All subvolumes are down" when all bricks are online
Keywords:
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: GlusterFS
Classification: Community
Component: replicate
Version: mainline
Hardware: x86_64
OS: Linux
unspecified
medium
Target Milestone: ---
Assignee: Karthik U S
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2019-06-20 08:55 UTC by ryan
Modified: 2022-02-08 04:19 UTC (History)
4 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2022-02-08 04:19:40 UTC
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Embargoed:


Attachments (Terms of Use)
Samba client log (2.30 MB, text/plain)
2019-06-20 08:55 UTC, ryan
no flags Details
Brick log (11.34 KB, text/plain)
2019-06-20 08:55 UTC, ryan
no flags Details
MCV01 volume info (1.88 KB, text/plain)
2022-02-07 11:42 UTC, ryan
no flags Details
MCV01 client volfiles (7.70 KB, application/zip)
2022-02-07 11:42 UTC, ryan
no flags Details

Description ryan 2019-06-20 08:55:10 UTC
Created attachment 1582578 [details]
Samba client log

Description of problem:
Gluster VFS client logs print ' E [MSGID: 108006] [afr-common.c:5413:__afr_handle_child_down_event] 0-mcv01-replicate-8: All subvolumes are down. Going offline until at least one of them comes back up.' repeatedly. When checking 'gluster vol status' all bricks are online. There are no obvious errors in the glusterd logs. The errors go back months, from when we had a failed storage array, where all the bricks had to be removed and replaced.

Volume is a distributed-replicated (x2) type.

In a few of the brick logs, there are a few RPC errors.
Other than the logs being flooded with this message, the cluster seems to be operating normally.
Iperf tests between the nodes does not flag any issues, with 0 TCP retries on all.


Version-Release number of selected component (if applicable):
Gluster 4.1.8
Samba 4.9.4

Comment 1 ryan 2019-06-20 08:55:32 UTC
Created attachment 1582579 [details]
Brick log

Comment 2 Atin Mukherjee 2019-06-25 09:19:01 UTC
Could you paste the gluster v status output? Are all bricks running fine? The error logs indicate that client is unable to connect to the bricks. Have you checked if any firewall settings blocking your client to connect to the bricks?

Comment 3 ryan 2019-06-25 09:30:21 UTC
Hi Atin,

Here's the output you requested:
I believe the issue is from when we had a lot of bricks (50%) which needed replacing, and then syncing with 'gluster volume sync', and it seems to be remembering something.

Status of volume: ctdbv01
Gluster process                             TCP Port  RDMA Port  Online  Pid
------------------------------------------------------------------------------
Brick mcn01:/mnt/ctdb/data                  49157     0          Y       6118
Brick mcn02:/mnt/ctdb/data                  49157     0          Y       3294
Brick mcn03:/mnt/ctdb/data                  49181     0          Y       49199
Brick mcn04:/mnt/ctdb/data                  49181     0          Y       194533
Self-heal Daemon on localhost               N/A       N/A        Y       70714
Self-heal Daemon on mcn04                   N/A       N/A        Y       146680
Self-heal Daemon on mcn03                   N/A       N/A        Y       141658
Self-heal Daemon on mcn01                   N/A       N/A        Y       147639

Task Status of Volume ctdbv01
------------------------------------------------------------------------------
There are no active volume tasks

Status of volume: ctdbv02
Gluster process                             TCP Port  RDMA Port  Online  Pid
------------------------------------------------------------------------------
Brick mcn01:/mnt/ctdb/dmz_data              49158     0          Y       6126
Brick mcn02:/mnt/ctdb/dmz_data              49158     0          Y       3302
Brick mcn03:/mnt/ctdb/dmz_data              49182     0          Y       49570
Brick mcn04:/mnt/ctdb/dmz_data              49182     0          Y       194825
Self-heal Daemon on localhost               N/A       N/A        Y       70714
Self-heal Daemon on mcn04                   N/A       N/A        Y       146680
Self-heal Daemon on mcn03                   N/A       N/A        Y       141658
Self-heal Daemon on mcn01                   N/A       N/A        Y       147639

Task Status of Volume ctdbv02
------------------------------------------------------------------------------
There are no active volume tasks

Status of volume: dmzv01
Gluster process                             TCP Port  RDMA Port  Online  Pid
------------------------------------------------------------------------------
Brick mcn01:/mnt/h1a/dmz_data               49179     0          Y       7495
Brick mcn03:/mnt/h1b/dmz_data               49183     0          Y       54973
Brick mcn01:/mnt/h2a/dmz_data               49180     0          Y       7614
Brick mcn03:/mnt/h2b/dmz_data               49184     0          Y       54996
Brick mcn01:/mnt/h3a/dmz_data               49181     0          Y       7803
Brick mcn03:/mnt/h3b/dmz_data               49185     0          Y       55019
Brick mcn01:/mnt/h4a/dmz_data               49182     0          Y       7936
Brick mcn03:/mnt/h4b/dmz_data               49186     0          Y       55042
Brick mcn01:/mnt/h5a/dmz_data               49183     0          Y       8169
Brick mcn03:/mnt/h5b/dmz_data               49187     0          Y       55065
Brick mcn02:/mnt/h6a/dmz_data               49159     0          Y       3312
Brick mcn04:/mnt/h6b/dmz_data               49179     0          Y       2924
Brick mcn02:/mnt/h7a/dmz_data               49160     0          Y       3321
Brick mcn04:/mnt/h7b/dmz_data               49180     0          Y       2947
Brick mcn02:/mnt/h8a/dmz_data               49161     0          Y       3329
Brick mcn04:/mnt/h8b/dmz_data               49183     0          Y       2970
Brick mcn02:/mnt/h9a/dmz_data               49162     0          Y       3338
Brick mcn04:/mnt/h9b/dmz_data               49184     0          Y       2998
Brick mcn02:/mnt/h10a/dmz_data              49163     0          Y       3346
Brick mcn04:/mnt/h10b/dmz_data              49185     0          Y       3024
Self-heal Daemon on localhost               N/A       N/A        Y       70714
Quota Daemon on localhost                   N/A       N/A        Y       70730
Self-heal Daemon on mcn03                   N/A       N/A        Y       141658
Quota Daemon on mcn03                       N/A       N/A        Y       141670
Self-heal Daemon on mcn04                   N/A       N/A        Y       146680
Quota Daemon on mcn04                       N/A       N/A        Y       146691
Self-heal Daemon on mcn01                   N/A       N/A        Y       147639
Quota Daemon on mcn01                       N/A       N/A        Y       147654

Task Status of Volume dmzv01
------------------------------------------------------------------------------
There are no active volume tasks

Status of volume: mcv01
Gluster process                             TCP Port  RDMA Port  Online  Pid
------------------------------------------------------------------------------
Brick mcn01:/mnt/h1a/mcv01_data             49174     0          Y       6690
Brick mcn03:/mnt/h1b/mcv01_data             49157     0          Y       54637
Brick mcn01:/mnt/h2a/mcv01_data             49175     0          Y       6842
Brick mcn03:/mnt/h2b/mcv01_data             49158     0          Y       54663
Brick mcn01:/mnt/h3a/mcv01_data             49176     0          Y       6936
Brick mcn03:/mnt/h3b/mcv01_data             49159     0          Y       54689
Brick mcn01:/mnt/h4a/mcv01_data             49177     0          Y       7153
Brick mcn03:/mnt/h4b/mcv01_data             49160     0          Y       54715
Brick mcn01:/mnt/h5a/mcv01_data             49178     0          Y       7303
Brick mcn03:/mnt/h5b/mcv01_data             49161     0          Y       54741
Brick mcn02:/mnt/h6a/mcv01_data             49164     0          Y       3358
Brick mcn04:/mnt/h6b/mcv01_data             49154     0          Y       119700
Brick mcn02:/mnt/h7a/mcv01_data             49165     0          Y       3366
Brick mcn04:/mnt/h7b/mcv01_data             49152     0          Y       165703
Brick mcn02:/mnt/h8a/mcv01_data             49169     0          Y       3608
Brick mcn04:/mnt/h8b/mcv01_data             49153     0          Y       37527
Brick mcn02:/mnt/h9a/mcv01_data             49170     0          Y       3616
Brick mcn04:/mnt/h9b/mcv01_data             49160     0          Y       2746
Brick mcn02:/mnt/h10a/mcv01_data            49171     0          Y       3624
Brick mcn04:/mnt/h10b/mcv01_data            49161     0          Y       2773
Self-heal Daemon on localhost               N/A       N/A        Y       70714
Quota Daemon on localhost                   N/A       N/A        Y       70730
Self-heal Daemon on mcn03                   N/A       N/A        Y       141658
Quota Daemon on mcn03                       N/A       N/A        Y       141670
Self-heal Daemon on mcn04                   N/A       N/A        Y       146680
Quota Daemon on mcn04                       N/A       N/A        Y       146691
Self-heal Daemon on mcn01                   N/A       N/A        Y       147639
Quota Daemon on mcn01                       N/A       N/A        Y       147654

Task Status of Volume mcv01
------------------------------------------------------------------------------
Task                 : Rebalance
ID                   : a4cabb46-efb1-4909-b059-c115d811b519
Status               : completed


Best,
Ryan

Comment 4 Sanju 2019-07-11 02:28:29 UTC
@Karthik, can you please look into this?

Comment 5 ryan 2019-09-17 14:17:40 UTC
Hi all,

Is there anything I can do to progress this?

Best,
Ryan

Comment 6 Sanju 2019-09-18 07:35:28 UTC
Hi Karthik,

I'm moving this bug to afr as we don't have any issue from glusterd here. If anything is needed from glusterd side, feel free to reach out.

Thanks,
Sanju

Comment 7 Karthik U S 2019-09-19 09:56:58 UTC
Hi,

Sorry for the delay. Can you please answer/provide the following?
- Is this problem seen only with volume "mcv01"?
- Logs are showing that bricks on nodes "mcn01" & "mcn03" are down. There are other volumes with bricks on these nodes. Are they not flooding the client logs with similar messages?
- Are you able to do IO from this client without any errors?
- Whether the bricks which needed replacing were on these nodes and are they the first 10 bricks of volume mcv01?
- Provide the output of "gluster volume info mcv01".
- Provide the statedump of the client processes which are showing these messages. (https://docs.gluster.org/en/latest/Troubleshooting/statedump/#generate-a-statedump)
- Give the client vol file which will be present inside "/var/lib/glusterd/vols/<volume-name>/"

Regards,
Karthik

Comment 8 Karthik U S 2019-12-13 09:29:26 UTC
Since there is no update on this bug for almost 3 months and also the branch 4.1 is EOL, I am closing this bug for now. If this issue is still seen on any of the maintained branches, please feel free to re-open this or file a new bug with the gluster logs and all the information requested in comment #7.

Comment 9 ryan 2022-02-07 10:20:21 UTC
Hi @ksubrahm

Comment 10 ryan 2022-02-07 11:40:58 UTC
Hi @ksubrahm,

Sorry, I didn't see a notification from this bug regarding the update.
We're still seeing this issue with Gluster 8.5.

Please find the requested information below:
- Is this problem seen only with volume "mcv01"?
MCV01 is the only volume that we have exported via Samba, the other volume is just for CTDB. We've seen this on other clusters that have 2x2 replicated volumes though.

- Logs are showing that bricks on nodes "mcn01" & "mcn03" are down. There are other volumes with bricks on these nodes. Are they not flooding the client logs with similar messages?
The logs say this, but as far as I can tell, all bricks in the volume are healthy and online. I'm not seeing these messages via the FUSE client logs. We have MCV01 mounted on all nodes via FUSE, however clients never access the volume via this as we use the VFS module.
 
- Are you able to do IO from this client without any errors?
Yes, the system is fully functional we can read and write from the volume as expected. Replication looks to be working fine too, as there are no pending heals etc.

- Whether the bricks which needed replacing were on these nodes and are they the first 10 bricks of volume mcv01?
It seems that all replicated sub-volumes on all nodes are being effected.
[2022-02-07 10:28:15.890242] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-0: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.890549] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-1: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.890771] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-3: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.890933] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-4: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891049] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-2: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891225] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-5: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891355] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-6: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891528] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-7: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891655] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-8: All subvolumes are down. Going offline until at least one of them comes back up.
[2022-02-07 10:28:15.891788] E [MSGID: 108006] [afr-common.c:6071:__afr_handle_child_down_event] 0-mcv01-replicate-9: All subvolumes are down. Going offline until at least one of them comes back up.

- Provide the output of "gluster volume info mcv01".
Please find this uploaded to the ticket with filename mcv01_info_07022022.txt

- Provide the statedump of the client processes which are showing these messages. (https://docs.gluster.org/en/latest/Troubleshooting/statedump/#generate-a-statedump)
Could you advise how to get a statedump from a VFS client? I tried the usual way but the dump was not being generated.
 
- Give the client vol file which will be present inside "/var/lib/glusterd/vols/<volume-name>/"
There were quite a few client volfiles. I've compressed them and attached them to this ticket with the filename mcv01_client_volfiles_07022022.zip.

I did have one thought: Could the 'auth.allow: 172.30.30.*' be at fault here? We use that to prevent non-cluster nodes from mounting the volume, but I'm wondering if we need to include 127.0.0.1 and other local loopback addresses in this, as the Samba server is installed on all Gluster nodes, so it may be that the connection is coming over a local loopback IP rather than their backend network IP.

Comment 11 ryan 2022-02-07 11:42:02 UTC
Created attachment 1859553 [details]
MCV01 volume info

Comment 12 ryan 2022-02-07 11:42:32 UTC
Created attachment 1859554 [details]
MCV01 client volfiles

Comment 13 Karthik U S 2022-02-08 04:19:40 UTC
Hi Ryan,

Thank you for reaching out on this bug. We have moved from bugzilla to github issues sometime back.
Glusterfs-8 is EOL and the maintained branches are Glusterfs-9 & 10 [1].
Please upgrade to a maintained version and if you still face the same, please file a new issue on github [2] with all the necessary details.

(Reverting the bug status back to closed.)

[1] https://www.gluster.org/release-schedule/
[2] https://github.com/gluster/glusterfs/issues/new

Regards,
Karthik


Note You need to log in before you can comment on or make changes to this bug.