[LU-3098] failure to auto trigger paused LFSCK causes mount to fail on slave MDTs

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.4.0
• Affects Version/s: Lustre 2.4.0
• Labels: lfsck mdt
• Severity: 3
• Assignee: WC Triage
• Reporter: John Hammond

Description

After 7e915f5d, in mdt_prepare() if the OBD_IOC_START_LFSCK ioctl() fails and mdt->mdt_seq_site.ss_node_id != 0 then the MDT will fail to mount.

static int mdt_prepare(const struct lu_env *env,
                struct lu_device *pdev,
                struct lu_device *cdev)
{
	...

	rc = mdt->mdt_child->md_ops->mdo_iocontrol(env, mdt->mdt_child,
                                                   OBD_IOC_START_LFSCK,
                                                   0, NULL);
        if (rc != 0)
                CWARN("Fail to auto trigger paused LFSCK.\n");

	if (mdt->mdt_seq_site.ss_node_id == 0) {
                rc = mdt->mdt_child->md_ops->mdo_root_get(env, mdt->mdt_child, &mdt->mdt_md_root_fid);
                if (rc)
                        RETURN(rc);
        }

        ...
}

If this is really the intended behavior (I assume not) then the CWARN() should be upgraded to CERROR().

Comments

[Jodi Levi (Inactive)]

Patch landed. If more work is needed let me know and I'll reopen this ticket.

[John Hammond]

Please see http://review.whamcloud.com/5931.