Bug 1568581 - iSCSI: Add warning about restarting tcmu-runner manually
Summary: iSCSI: Add warning about restarting tcmu-runner manually
Alias: None
Product: Red Hat Ceph Storage
Classification: Red Hat
Component: Documentation
Version: 3.0
Hardware: Unspecified
OS: Unspecified
Target Milestone: z2
: 3.0
Assignee: Aron Gunn
QA Contact: Madhavi Kasturi
Depends On:
TreeView+ depends on / blocked
Reported: 2018-04-17 19:59 UTC by Mike Christie
Modified: 2019-08-26 06:55 UTC (History)
4 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Last Closed: 2019-08-26 06:55:42 UTC
Target Upstream Version:

Attachments (Terms of Use)

Description Mike Christie 2018-04-17 19:59:23 UTC
Description of problem:

If tcmu-runner is not running and a initiator sends a command, the command will be stuck until you reboot the box. The stuck command will force the initiator's error handler to run, the LUN to be offlined and in windows/linux the session will be dropped and future logins will fail.

Normally, the user should not be starting/stopping tcmu-runner manually. Components like rtslib and rbd-target-gw/ceph-iscsi-config and systemd will handle this.

However, users will sometimes start/stop services as part of their debugging, so this is a request to add a warning in the release notes or main doc like here


or wherever something like this should go.

The warning should be something like:

tcmu-runner must not be stopped while iSCSI initiators are logged into the target. Before stopping tcmu-runner, the rbd-target-gw service must be stopped cleanly, or the iSCSI initiators must be logged out. If this is not done, SCSI commands may get stuck in the target layer and the system will have to be rebooted.

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:

Actual results:

Expected results:

Additional info:

A kernel fix is required and that will be added to 7.6. A bz for that is here


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