HP StorageWorks Scalable File Share Release Notes - Version 2.3

Known issues and workarounds in HP SFS Version 2.3 1–9
If your XC systemimage is standard/unedited, you can simply remove the current systemimage contents and
re-create a new golden image using the following commands:
With the systemimage suitably updated, you can now reimage your clients as described in the XC
documentation.
1.4.6 First boot of XC compute nodes may hang or fail during SFS nconfig
operation
When booting freshly imaged XC compute nodes, especially if the /hptc_cluster filesystem is served
by SFS, the C10hptc_cluster_fs nconfig operation could hang or fail.
This is sometimes related to difficulty running sfsconfig within the C10hptc_cluster_fs nconfig
step during the first boot following imaging. The sfsconfig operation requires the client to have network
connectivity to the SFS server alias. On an XC, this is typically achieved by routing via the client's assigned
nat/external server. However, there are times when either sfsconfig fails to contact the SFS server, or
sfsconfig fails for other reasons.
Initial troubleshooting includes reviewing the /etc/modprobe.conf(.lustre) contents within the
golden image on the XC headnode, then repeating the startsys image_and_boot operation to see if
the nconfig/sfsconfig operation is successful.
If unsuccessful, then it could help to monitor the serial console logs during the boot operation. However, to
fully diagnose the problem, you might need to cancel the startsys image_and_boot operation while
the nconfig is still running, and then connect to the client's serial console. Canceling the startsys will
keep the node booted and therefore allow deeper investigation of the cause of the problem, such as by
running sfsconfig -X on the console and observing the output. Also, while logged into the client, you
can inspect the contents of /var/log/messages, and compare the contents of the /etc/
modprobe.conf(.lustre) files with the successful environment of the XC headnode.
1.4 . 7 lfs quota ch e ck issu es
When running lfs quotacheck, specify only one option—either -u or -g. By not specifying an option,
or by specifying both, lfs quotacheck could hang. Allow an lfs quotacheck command adequate
time to complete. Additionally, when running lfs quotacheck, if the client process is ended, this can
cause a server LBUG with the following assertion: ASSERTION(imp->imp_obd != NULL)
1.4.8 colplot MDS performance counters do not reflect effective metadata activity
The SFS colplot web interface displays MDS traffic using /proc counters named mds_close,
mds_getattr, mds_reint and mds_sync. Lustre 1.4.11 (the Lustre version used in SFSV2.3) introduced
new /proc counters to report MDS activity, such as mds_getxattr, and mds_readpage. The
colplot/collectl package does not process these counters. The consequence is that the MDS counters
displayed by colplot are not an accurate indicator of the metadata activity.
Command Purpose
/usr/bin/si_lsimage
List available images
/usr/sbin/si_rmimage base_image
Delete the base_image
/opt/hptc/sbin/updateimage --gc n0 --init
Create a completely new base_image