Negative FC HBA Counters?

[Ralph Grothe]

Hi,

I want to write a wee Munin plug-in to collect and chart disk i/o for SAN storage.

The only useful tool I have discovered so far is one of fcmsutil|tdutil.
Though I would rather prefer to glean byte rates there only seem to be stats of so called "Assisted IOs".

Ok, but here I've got a negative counter on one of the controllers.
Does this simply mean a counter overflow (viz. data type excess), or does it imply some sort of direction (viz. reads vs. writes)?
But the latter seems nonsense to me.

$ /opt/fcms/bin/tdutil /dev/td0|grep IOs
Number of Assisted IOs = -1195641050
$ /opt/fcms/bin/tdutil /dev/td1|grep IOs
Number of Assisted IOs = 25072614


Also, if I fetch the devstats I get three counters of assisted IOs, each associated to another Nport.

$ /opt/fcms/bin/tdutil /dev/td0 devstat all|grep -e Nport -e Assisted
Device Statistics for Nport_id 0x654713
Assisted I/O requests 97838344
Device Statistics for Nport_id 0x655313
Assisted I/O requests 2911343631
Device Statistics for Nport_id 0x6e1113
Assisted I/O requests 92423487


But unfortunately, non of these Nports seems to be the one of the local controller according to deviating addresses.

$ /opt/fcms/bin/tdutil /dev/td0 devstat all|grep Nport
Device Statistics for Nport_id 0x654713
Device Statistics for Nport_id 0x655313
Device Statistics for Nport_id 0x6e1113
$ /opt/fcms/bin/tdutil /dev/td0|grep -i local
Local N_Port_id is = 0x655213


Rgds
Ralph

[Ralph Grothe]

Addendum, if my negative counter is owe to an overrun, was it than advisable to flush the stats by issuing as root something like

/opt/fcms/bin/fcmsutil /dev/td0 clear_stat

or would this have any unwanted side effects?