Lustre 1.6 comes with a series of new utilities to help configure and
analyze Lustre filesystems. These are all in the 1.6 betas, but have
been unadvertised. Hopefully they will be useful; further suggestions
are welcome -- just reply here.
A couple of notes
* these are not in the 1.6 manual yet
* the profiling tools won't in general work with liblustre; we are
working on that for a later 1.6.x release.
1. General purpose
* /usr/sbin/lustre_rmmod.sh
Remove all Lustre and LNET modules, assuming no Lustre services are
running (safe).
2. Managing large clusters
* /usr/sbin/lustre_config.sh
helps automate the formatting and setup of disks on multiple
nodes. An entire installation is described in a comma-separated
file and passed to this script, which then formats the drives,
updates modprobe.conf, and produces HA configuration files.
* /usr/sbin/lustre_createcsv.sh
generates a CSV file describing a currently-running installation.
* /usr/sbin/lustre_up14.sh
grabs client configuration files from old MDT's. When upgrading a
1.4.x Lustre system to 1.6, if the MGS is not co-located with the
MDT or the client name is non-standard, this utility is used to
retrieve the old client log. (See
https://mail.clusterfs.com/wikis/lustre/UpgradeTo16)
3. Application profiling
* /usr/bin/lustre_req_history.sh
Run from a client, this script assembles as much Lustre RPC request
history as possible from the local node and from the servers that
were contacted, providing a better picture of the coordinated
network activity.
* /usr/bin/llstat.sh
Improved llstat handles a wider range of proc files, and has command
line switches to produce more graphable output.
* /usr/bin/plot-llstat.sh
Plots the output from llstat using gnuplot
4. More /proc statistics for application profiling.
* client vfs_ops_stats
Track Linux VFS operation calls into Lustre for a single PID, PPID,
GID, or everything. (bz10968)
/proc/fs/lustre/llite/*/vfs_ops_stats
/proc/fs/lustre/llite/*/vfs_track_[pid|ppid|gid]
* client extents_stats
Show the size distribution of I/O calls from the client, cumulative
and by process. (RT22484)
/proc/fs/lustre/llite/*/extents_stats, extents_stats_per_process
* client offset_stats
Show the read/write seek activity of a client by offsets and ranges.
(RT22484)
/proc/fs/lustre/llite/*/offset_stats
* per-client stats tracked on the servers
Each MDT and OST now tracks LDLM and operations stats for every
connected client, for comparisons or simpler collection of
distributed job stats. (RT22485)
/proc/fs/lustre/mds|obdfilter/*/exports/
* Finer MDT stats
More detailed MDT operations stats are collected for better
profiling (RT22486) /proc/fs/lustre/mds/*/stats
5. Testing / debugging
* /usr/bin/loadgen
A test program used to generate large loads on local or remote OSTs
or echo servers. See https://mail.clusterfs.com/wikis/lustre/LoadGen
* /usr/sbin/llog_reader
Translate a Lustre configuration log into human-readable form.
* /usr/sbin/lr_reader
Translate a last_rcvd file into human-readable form
_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
