1 | =============================================================================== |
---|
2 | filehist - Shell script wrapper for generating file space histograms |
---|
3 | for all currently mounted GPFS filesystems |
---|
4 | (uses /usr/lpp/mmfs/samples/util/tsinode |
---|
5 | and /usr/lpp/mmfs/samples/debugtools/tsfilehist.awk) |
---|
6 | Generates one file for each filesystem called $fsname.filesum |
---|
7 | |
---|
8 | tsfilehist.awk - Awk script for generating file space histograms on GPFS |
---|
9 | filesystems. (Used by filehist script) |
---|
10 | |
---|
11 | =============================================================================== |
---|
12 | trsum.awk - Awk script for analyzing a GPFS trace and reporting various |
---|
13 | statistics. |
---|
14 | |
---|
15 | Example to show application calls and IO performed. |
---|
16 | (The trace will wrap and only retain the records just |
---|
17 | before the stop. See mmtrace -? for other options to |
---|
18 | make the trace output larger.) |
---|
19 | |
---|
20 | mmtrace trace=io |
---|
21 | #run test program |
---|
22 | mmtrace stop #produces file with prefix /tmp/mmfs/trcrpt. |
---|
23 | cd /tmp/mmfs |
---|
24 | for file in trcrpt.*; do |
---|
25 | echo $file |
---|
26 | trsum.awk $file > trsum.${file#*trcrpt.} |
---|
27 | done & |
---|
28 | |
---|
29 | The trsum output shows system call entry/exit and elapsed time |
---|
30 | along with arguments for each operation. |
---|
31 | (trsum.awk details=0 ... to turn this part off) |
---|
32 | |
---|
33 | At the end of the report are summaries and performance histograms. |
---|
34 | |
---|
35 | =============================================================================== |
---|
36 | fsstruct.awk - Error report analysis on AIX clusters. |
---|
37 | Build a file with the names of all nodes in the GPFS cluster: |
---|
38 | mmcommon getNodeList|awk '{print $3}' > $WCOLL |
---|
39 | Gather errpt -a output from all nodes |
---|
40 | for node in $(<$WCOLL); do |
---|
41 | mmdsh -L $node errpt -a 2>/dev/null |
---|
42 | done > errpt.all & |
---|
43 | Then summarize the errpt.all file with one line per event: |
---|
44 | fsstruct.awk errpt.all | sort -k1,1 > gpfs.err.summary |
---|
45 | fsstruct.awk all=1 errpt.all | sort -k1,1 > all.err.summary |
---|
46 | |
---|
47 | =============================================================================== |
---|
48 | fsstructlx.awk - Error report analysis on Linux clusters. |
---|
49 | Build a file with the names of all nodes in the GPFS cluster: |
---|
50 | mmcommon getNodeList|awk '{print $3}' > $WCOLL |
---|
51 | Gather /var/log/messages* from all nodes |
---|
52 | for node in $(<$WCOLL); do |
---|
53 | mmdsh -L $node 'cat $(ls -tr /var/log/messages*)' 2>/dev/null |
---|
54 | done > syslogs.all & |
---|
55 | Then summarize the syslogs.all file with one line per event: |
---|
56 | fsstructlx.awk syslogs.all | sort -k1,1 > gpfs.err.summary |
---|
57 | fsstructlx.awk all=1 syslogs.all | sort -k1,1 > all.err.summary |
---|
58 | |
---|
59 | =============================================================================== |
---|