Resize of volume with a Data Change Object (DCO) hangs and also hangs vxconfigd on AIX

book

Article ID: 100003375

calendar_today

Updated On:

Description

Error Message

 

 

 

No specific error resulted. At the time, the resize command hung for 40 minutes before it was killed. All Volume Manager-based commands were also hung.

Cause

 

 

 

Examination of the threads associated with the vxconfigd process revealed that one of them, associated with the resize of the volume, was held up in DCO code.

Resolution

 

 

 

In 5.1RP2 of Volume Manager, a tuning change is made to help avoid the hang. It raises volpagemod_max_memsz to 64MB. This should help avoid all such related vxconfigd hangs. If upgrade to 5.1RP2 cannot be made, the tunable can be raised dynamically, and the resize command rerun:

# vxtune volpagemod_max_memsz 64M


# vxtune
Tunable Value Description
----------------- -------- ---------------
vol_rvio_maxpool_sz 131072 RVIO Pool Size (KBytes)
vol_min_lowmem_sz 13108 Low Memory Threshold (KBytes)
vol_max_rdback_sz 131072 Readback Pool Size (KBytes)
vol_max_nmpool_sz 32768 NMCOM Pool Size (KBytes)
vol_max_wrspool_sz 16384 Write Shipping Pool Size (KBytes)
volpagemod_max_memsz 65536 Cache Size for FMR Metadata (KBytes) <<<<<<
voldrl_volumemax_drtregs 256 Per volume dirty regions in log-plex DRL (Units)
voldrl_volumemax_drtregs_20 1024 Per volume dirty regions in DCO version 20 (Units)
vol_stats_enable 1 Enable VM I/O stat collection
vol_max_siobuf_sz 40140 SIO Buffer Pool Size (KBytes)


Applies To

 

 

 

AIX 5.3

Volume Manager 5.1

Volume with DCO object attached

Issue/Introduction

              When attempting to resize a volume using Volume Manager's vxassist command, the resize operation hung, and, subsequently, all Volume Manager commands hung.