On Red Hat Enterprise Linux (RHEL) 6.6 the umount(8) command hangs if an application is watching for inodes events on a VxFS file system using inotify(7) APIs

On Red Hat Enterprise Linux (RHEL) 6.6 the umount(8) command hangs if an application is watching for inodes events on a VxFS file system using inotify(7) APIs

Article: 100014551
Last Published: 2015-02-12
Ratings: 0 0
Product(s): InfoScale & Storage Foundation

Problem

 

On Red Hat Enterprise Linux (RHEL) 6.6 additional counters were added in the superblock to track inotify watches, these new counters were not implemented in VxFS. Whilst doing a umount, the operation hangs until the counter in the superblock drops to zero, which will not happen as they are not handled in VxFS. This applies to both VxFS mounts and Cluster File Systems (CFS) mounts
This issue could prevent a cluster failover if the file system cannot be un-mounted.

Solution

Veritas have produced patches to resolve this issue, see the table below:

 

VxFS patches for inotify unmount issue patches
SFHA Version Patch
6.0.5

fs-rhel6_x86_64-6.0.5.200

6.1.1

fs-rhel6_x86_64-6.1.1.200

6.2

fs-rhel6_x86_64-6.2.0.100

Previously, a fix was available in the form of a private hot fix (fs-rhel6_x86_64-6.2.0.001)

 

Note: The patches above for 6.0.5 and 6.1.1 are also contained in the combined patches sfha-rhel6.6_x86_64-6.0.5.200 and sfha-rhel6.6_x86_64-6.1.1.200, If installing SFHA for the first time, it may be more efficient to use these combined patches (or later). 

Workaround (if no patch):

Abort the applications that are using inotify APIs (poll(2), select(2) and epoll(7)) before un-mounting the file system

Applies To

Storage Foundation (SF) 6.2, 6.1.1 & 6.0.5
Storage Foundation High Availability (SFHA) 6.2, 6.1.1 & 6.0.5
Storage Foundation Cluster File System High Availability (SFCFSHA) 6.2, 6.1.1 & 6.0.5
Storage Foundation Cluster for Oracle RAC (SFRAC) 6.2, 6.1.1 & 6.0.5
VxFS 6.2, 6.1.1 & 6.0.5
Red Hat Enterprise Linux (RHEL) 6.6

 

Was this content helpful?