How many data (MB/GB) go through the HBA-card

[Rob van Buiten]

I have a unix server (rp7420) with HP-UX 11.11
This server have 2 HBA cards and I want to know the throughput of each card.

underlying question:
every day a backup takes about 1,5 hours only sunday at takes 4/5 hours.

so, I want to monitor

Paul.

[Michal Kapalka (mikap)]

hi,

check the HBA :

ioscan -fnC fc

fcmsutil /dev/HBA_CARD vpd

mikap

[Andrew Rutter]

hi,

is it the same backup you are running? is it not just a full backup on a sunday and an incremental backup daily?

If its the same i would look at not just the HBA horughput, you could do with using glance and see what else the system is doing

Andy

[Rob van Buiten]

@Mikap
output is:
root# fcmsutil /dev/td0 vpd
V I T A L P R O D U C T D A T A
--------- ------------- -------

Product Discription :'A6795A 2Gbps/1Gbps Fibre Channel 4X PCI HBA'
Part number :'A6795-62002'
Engineering Date Code :'B-4524'
Part Serial number :'A56A00348784'
Misc. Information :'PW=15W'
Mfd. Date :'A-4528'
Check Sum : 0x3f
EFI version : '001.17'
Asset Tag :'NA'

But wat is using?

[Rob van Buiten]

@Andy,

Yes, it is the same backup (always a full oracle backup)
more info, this full backup is started every 3 hours, and all backups (except 1) is ready < 2 hours normaly 1,5)


Paul.

[Bill Hassell]

Are you backing up from one disk to another disk? (if you're writing to tape, then there are a number of additional settings to configure). Is your backup specification the same for every day? Are you performing any other system activity during the backup period?

[Rob van Buiten]

@Bill and others,

Backup is from oracle to disk (san). And the same backup every 3 hours.
Iâ m aware that this is complex, because also the DP-cell-manager is involved (IDB).
What I want to know (to log) what really the throughput is at every backup time.
And from every HBA card.

What I see in DP rman session log is.
It takes (at problem backup) more time (minutes) to backup 1 oracle datafile.
So, I want (also) to know the throughput of each HBA.


Paul.

[Rita C Workman]

Before I'd go through checking that I would check to see if the DBA's are doing something then....like maybe exports.

I've seen this kind of happening on backups, where the DBA's started doing exports and other massive copying of a couple databases (for dev/testing) at the same time the backup process was running and hitting the same DB/same disk.

Just a thought,
Rita

[Alzhy]

First to answer your question:

You seem to have the classic older Tachyon XL2 Cards. Card is capabale of 2Gbps or 1Gbps. Check link speed via (fcmsutil /dev/tdX}grep -i "link speed"). A 2Gbps link will give you a maximum thrpughput of 180 MEGABYTEs/sec, a 1Gbps - of course about half that. But realistically -- these throughputs are rarely sustained as the bottleneck is usually the RAID layout on your SAN Array.


Now to your other question of why you daily backups takelonger on a weekend than on a weekday. Assuming the scripts used are the same, RMAN streams per channel are the same too -- then I can only surmise than on a Sunday -- your destination SAN disk for your backups is likely serving other disk i/o intensive activities too.

What kind of SAN disks are involved if you don't mind? Do you administer your SAN Array yourself or a different human?