Operating System - OpenVMS
1828204 Members
2289 Online
109975 Solutions
New Discussion

Re: IO checking on an HBA/fibre card in an ES40

 
SOLVED
Go to solution
Kirk Reindl
Frequent Advisor

IO checking on an HBA/fibre card in an ES40

System specs:
OVMS 2-node cluster
ver 7.3-1
EMC Symmetric 3930 disk array

I'd like to gather either some real-time or historical IO stats specifically on the HBAs/fibre cards in my systems.

I realize I can use show dev /multi pgx to examine IO Operations. But my IO completions doesn't appear to be dynamic here.

OSIJX1> show dev /full pgc

Device PGC0:, device type SCSI FCP, is online, error logging is enabled.

Error count 0 Operations completed 188
Owner process "" Owner UIC [SYSTEM]
Owner process ID 00000000 Dev Prot S:RWPL,O:RWPL,G,W
Reference count 0 Default buffer size 65535
Current preferred CPU Id 1 Fastpath 1

OSIJX1>

I'm quite sure I have IO being completed down the "pgc" channel. Look at the output from this command as I interrogate a disk whose current path is down PGC(I didn't supply all the output, just what I thought was relevant--I ran this command three times approximately 30 seconds apart from one another):

OSIJX1> show dev dga357 /full

Path PGC0.5006-0482-BCC3-0A5A (OSIJX1), current path.
Error count 0 Operations completed 299217321

Path PGC0.5006-0482-BCC3-0A5A (OSIJX1), current path.
Error count 0 Operations completed 299232886

Path PGC0.5006-0482-BCC3-0A5A (OSIJX1), current path.
Error count 0 Operations completed 299249797

It seems strange to me that I wouldn't see Operations completed increase on the HBA when I run,

OSIJX1> show dev /full pgc
Everything I run that command it is staying at,
Operations completed 188

Can someone tell me why this might be?
Also, we are running the Polycenter Capacity planning and performance analyzer product.
Does any one know of a way that I can use either this product or any of the monitor commands to drill down on the IO on any given HBA?

Thank you,
Kirk



3 REPLIES 3
Uwe Zessin
Honored Contributor

Re: IO checking on an HBA/fibre card in an ES40

Hello Kirk,
it looks to me that the PG device on OpenVMS are responsible for the FC-4 mapping. The adapter hardware appears to be represented by 'FG' devices. I have checked a log from a workshop I gave two weeks ago that used a pretty much current V7.3-2 system and it didn't report any operations at all on FG devices.

A fibre channel adapter has to do a bit more than just sending I/Os to a target and receiving responses. It is possible that nobody bothered to define what counts as an 'operation' or even to maintain a counter.
.
Volker Halle
Honored Contributor

Re: IO checking on an HBA/fibre card in an ES40

Kirk,

there don't seem to be any transfer or operations related counters inside either PG or FG data structures. It's the same as with all other SCSI port drivers (PKx).

If you really want to sum up the operations across all your PG/FG FC adapters, consider to parse the output of SHOW DEV/FULL DG and sum up all Operations Completed counters for each of the pathes via each adapter. The Operations Completed count for a multipath SCSI disk seems to be the sum of all operations completed on all pathes of that disk.

Maybe you can look at counters on the FC switch ?

Volker.
Uwe Zessin
Honored Contributor
Solution

Re: IO checking on an HBA/fibre card in an ES40

A fibre channel switch deals with fibre channel frames. An I/O can consist of several frames.

While a frame has a max. size of 2K, the standard allows for transfers of up to 16 MegaBytes in one big I/O.
.