SCSI errors resulting in down Object Storage Daemons (OSDs)

This document (7023562) is provided subject to the disclaimer at the end of this document.

Environment

SUSE Linux Enterprise Server 12 Service Pack 3 (SLES 12 SP3)

SUSE Enterprise Storage 5

LSI MPT Fusion SAS 3.0 Device Driver

tuned

Situation

With the default driver version "15.100.00.00", SCSI commands are timing out resulting in the following SCSI errors and SUSE Enterprise Storage (SES) errors being seen:

2018-12-01T05:01:20.649519+01:00 server_name kernel: [40811.416245] sd 0:0:X:0: timing out command, waited 180s
2018-12-01T05:01:20.649528+01:00 server_name kernel: [40811.416279] sd 0:0:X:0: [sdX] tag#8 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
2018-12-01T05:01:20.649528+01:00 server_name kernel: [40811.416290] sd 0:0:X:0: [sdX] tag#8 Sense Key : Not Ready [current]

...

osd.xxx failed (root=xxxx,disktype=xxx,host=serverxxx) (xx reporters from different host after 100.964986 >= grace 100.287216)
we have enough reporters to mark osd.xxx down

...

map e422871 wrongly marked me down at ee422870

Resolution

The current workaround is to:

1. Make sure the latest HBA firmware is applied.

2. Update to the latest upstream mpt3sas driver version 26.00.00.00 available from the hardware vendor.

NOTE: See also the additional information section regarding "tuned".

Cause

Additional Information

Since SES 5.5 with DeepSea 0.8.6 we also are implementing and enabling tuned profiles. With the default tuned settings this results in power management being enabled for spinning disks. This can result in disks hosting OSDs to spin down, which is undesirable.

To disable power management on the drives hosting OSDs and to disable tuned for now take the following steps:

# hdparm -B 255 -S 0 /dev/sd<X>

# tuned-adm off
# systemctl stop tuned.service

To see the current disk status:

# hdparm -C /dev/sd<X>

Using "smartctl" which is part of the "smartmontools" package can also be useful, for example to view disk information (including "Start_Stop_Count"):

# smartctl -a /dev/sd<X>

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.