RHEL and Ubuntu Guest OS may report failure when fstrim command is triggered from OS side

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms:

• RHEL and Ubuntu Guest OS failure may be seen when fstrim command is triggered from OS side
• /var/log/messages

Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE cmd_age=0s
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] Sense Key : Vendor Specific(9) [current]
Oct 12 21:15:32 <name>kernel: sd 0:0:1:0: [sdb] <<vendor>>ASC=0x80 <<vendor>>ASCQ=0x85
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] CDB: Write same(16) 93 08 00 00 00 00 06 1e b2 58 00 00 20 00 00 00
Oct 12 21:15:32 <name> kernel: blk_update_request: I/O error, dev sdb, sector 102675032
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE cmd_age=0s
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] Sense Key : Vendor Specific(9) [current]
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] <<vendor>>ASC=0x80 <<vendor>>ASCQ=0x85
Oct 12 21:15:32 <name> kernel: sd 0:0:1:0: [sdb] CDB: Write same(16) 93 08 00 00 00 00 06 1e d2 58 00 00 20 00 00 00
Oct 12 21:15:32 <name> kernel: blk_update_request: I/O error, dev sdb, sector 102683224
Oct 12 21:15:32 <name> kernel: XFS (dm-8): discard failed for extent [0x2bdb0d,7], error -5

Environment

VMware vSphere ESXi 7.0

Cause

For certain workloads the Guest OS issued unmap can fail due to resource contention on ESXi side.
UNMAPs/ WRITE SAME issued by Guest OS need to lock a given resource cluster on the VMFS file system and if the same resource cluster is locked by a different work flow such as space allocation, then unmap request fails with resource contention.

Issue is not seen always and will be seen along with other workload that can cause resource contention for the unmap request.

Resolution

This issue is fixed on VMware ESXi 7.0 Update 3q - VMware ESXi 7.0 Update 3q Release Notes - (WRITE SAME (16) commands with the unmap bit set fail on thick disks).