The "extended" statistics provided by "vos examine -extended" are not
properly reset @ midnight.
That is, they are actually reset correctly, however "vos examine" does
not get the copy from the fileserver, but from the volume header. The
code there to deal with out-of-date statistics correctly sets the
number of accesses to 0, but leaves the extended statistics at their
old value. You can therefore end up with volumes of 0 accesses in the
past day and millions of writes.
Patch attached, Bcc'ed to openafs-bugs.
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rainer Toebbicke
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland
Phone: +41 22 767 8985 Fax: +41 22 767 7155
--- openafs/src/volser/volprocs.c.o144 2006-12-19 04:40:14.000000000 +0100
+++ openafs/src/volser/volprocs.c 2007-10-02 10:16:48.000000000 +0200
@@ -2052,11 +2052,6 @@
xInfoP->accessDate = volDiskDataP->accessDate;
xInfoP->updateDate = volDiskDataP->updateDate;
xInfoP->backupDate = volDiskDataP->backupDate;
- now = FT_ApproxTime();
- if (now - volDiskDataP->dayUseDate > OneDay)
- xInfoP->dayUse = 0;
- else
- xInfoP->dayUse = volDiskDataP->dayUse;
xInfoP->filecount = volDiskDataP->filecount;
xInfoP->maxquota = volDiskDataP->maxquota;
xInfoP->size = volDiskDataP->diskused;
@@ -2064,8 +2059,15 @@
/*
* Copy out the stat fields in a single operation.
*/
- memcpy((char *)&(xInfoP->stat_reads[0]),
+ now = FT_ApproxTime();
+ if (now - volDiskDataP->dayUseDate > OneDay) {
+ xInfoP->dayUse = 0;
+ memset((char *)&(xInfoP->stat_reads[0]), 0, numStatBytes);
+ } else {
+ xInfoP->dayUse = volDiskDataP->dayUse;
+ memcpy((char *)&(xInfoP->stat_reads[0]),
(char *)&(volDiskDataP->stat_reads[0]), numStatBytes);
+ }
/*
* We're done copying. Detach the volume and iterate (at this
