script/find_dups_in_pg_log: scrip to find dup requests due to short pg logs

Signed-off-by: Sage Weil <sage@redhat.com>

diff --git a/src/script/find_dups_in_pg_log.sh b/src/script/find_dups_in_pg_log.sh
new file mode 100755 (executable)
index 0000000..b4d1afb
--- /dev/null
+++ b/src/script/find_dups_in_pg_log.sh
@@ -0,0 +1,22 @@
+#!/bin/sh
+
+# pipe output of grep for objectname in osd logs to me
+#
+# e.g.,
+#
+#  zgrep smithi01817880-936 remote/*/log/*osd* | ~/src/ceph/src/script/find_dups_in_pg_log.sh
+#
+# or
+#
+#  zcat remote/*/log/*osd* | ~/src/ceph/src/script/find_dups_in_pg_log.sh
+#
+# output will be any requests that appear in the pg log >1 time (along with
+# their count)
+
+#grep append_log | sort -k 2 | sed 's/.*append_log//' | awk '{print $3 " " $8}' | sort | uniq | awk '{print $2}' | sort | uniq -c | grep -v ' 1 '
+
+grep append_log | grep ' by ' | \
+        perl -pe 's/(.*) \[([^ ]*) (.*) by ([^ ]+) (.*)/$2 $4/' | \
+        sort | uniq | \
+        awk '{print $2}' | \
+        sort | uniq -c | grep -v ' 1 '