[LU-12721] HPE MPT MPI and lustre2.12 clients causes write hangs

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.12.2
• Type: Bug
• Priority: Critical
• Reporter: Mahmoud Hanafi
• Assignee: Andreas Dilger
• Resolution: Duplicate
• Severity: 2

Description

Using HPE MPT MPI with lustre2.12 clients write hang for 30 sec.

Here is strace that show the timing. We can only write(1,"",0) hang. Looks like a duplicate of  LU-12382.

 

1567214852.343460 write(1, " Compile options:\n", 18) = 18
1567214852.343873 write(1, "", 0)       = 0
1567214883.059446 write(1, "    MPIF77       = ifort\n", 25) = 25
1567214883.059938 write(1, "", 0)       = 0
1567214913.783442 write(1, "    FLINK        = $(MPIF77)\n", 29) = 29
1567214913.783822 write(1, "", 0)       = 0
1567214944.499451 write(1, "    FMPI_LIB     = -lmpi\n", 25) = 25
1567214944.499933 write(1, "", 0)       = 0
1567214975.219462 write(1, "    FMPI_INC     = (none)\n", 26) = 26
1567214975.219942 write(1, "", 0)       = 0
1567215005.943478 write(1, "    FFLAGS       = -O3 -ipo -axC"..., 66) = 66
1567215005.944041 write(1, "", 0)       = 0
1567215036.663466 write(1, "    FLINKFLAGS   = $(FFLAGS) -lc"..., 48) = 48
1567215036.664046 write(1, "", 0)       = 0
1567215067.383452 write(1, "    RAND         = randi8\n", 26) = 26
1567215067.383861 write(1, "", 0)       = 0

Issue Links

• is related to: LU-12382 Deadloop on tiny write

Comments

[Andreas Dilger]

There is already a backported version of the LU-12382 patch https://review.whamcloud.com/35312/ that is landed to b2_12

[Peter Jones]

Mahmoud

Sounds like this is a duplicate ticket that will be fixed in 2.12.3

Peter