Project

General

Profile

Bug #37577

kclient mount does not detect newly active rank 0

Added by Patrick Donnelly 6 months ago. Updated 6 months ago.

Status:
New
Priority:
High
Assignee:
Category:
fs/ceph
Target version:
-
Start date:
Due date:
% Done:

0%

Source:
Development
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:

Description

First setting the cluster down and failing rank 0, then I started this mount:

time mount -t ceph $(grep 'mon host' /etc/ceph/ceph.conf | tr -d '[[:space:][:alpha:]=]'):/ /cephfs -o secret=$(grep key /etc/ceph/ceph.client.admin.keyring | awk '{print $3}'),name=admin,norequire_active_mds,mount_timeout=300
mount error 5 = Input/output error

real    10m0.036s
user    0m0.009s
sys     0m0.019s

Following that, after ~5 minutes, I set cluster_down=false. The mount did not detect rank 0 coming up and try to connect.

History

#1 Updated by Patrick Donnelly 6 months ago

The use of mount_timeout is confusing from a glance at the source code. For example, it's used to set the request timeout; I feel the request timeout should use a different mount option for that. The mount_timeout should be permitted to be infinite and the kernel should retry as necessary in the background.

Also available in: Atom PDF