This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: Issues with System Tap + iostat-scsi.stp + IBM Tape Drives


Hi! Thanks for the response.


Unfortunately the command I need to run is hidden behind menus or runs via an application (Tivoli Storage Manager).

Is there any way to open up tracing that just traces anything going on at the time? I don't have any one command that will generate tape traffic atm.

-----Original Message-----
From: Frank Ch. Eigler [mailto:fche@redhat.com] 
Sent: Wednesday, June 13, 2012 9:38 PM
To: Andy Wojnarek
Cc: systemtap@sourceware.org
Subject: Re: Issues with System Tap + iostat-scsi.stp + IBM Tape Drives


Hi, Andy -

andy.wojnarek wrote:

> I'm trying to use STAP and iostat-scsi to gather tape drive
> performance statistics on my tape drives. I can't get any
> information from tape drives though. [...]
> I am only getting disk stats: [...]
>
> ls -al /dev/tape/by-id
> lrwxrwxrwx 1 root root   9 Dec 27 13:05 scsi-1IBM_3573-TL_00L4U78P6926_LL0-changer -> ../../sg5
> lrwxrwxrwx 1 root root   9 Dec 27 13:05 scsi-32001000e111443b5-sg -> ../../sg4
> lrwxrwxrwx 1 root root   9 Dec 27 13:05 scsi-32004000e111443b5-sg -> ../../sg3

> If you could point me in the right direction I'd REALLY appreciate
> it. STAP is literally the only way I've found to get performance
> from scsi tape drives.

OK.  If the two probes in iostat-scsi.stp aren't catching this type of
tape drive traffic, try some tracing to figure out where the traffic
is going instead.

# stap .../para-callgraph.stp 'module("sg").function("*")' -c 'tape operation'
# stap .../para-callgraph.stp 'module("st").function("*")' -c 'tape operation'

to see what's up.  It may be that probing at a lower level, like
sg_start_req(), decoding the Sg_request struct (fields in srp->header)
can give the same device-name / transfer-size-direction data as the
sd_mod / st probes do.


> It's 2012 and there is still nothing easy that allows this to
> happen. On AIX, I can do iostat -p and it will show me the tape
> drives.

I get the sense more recent kernels have added some exposure of tape
drives to iostat or blktrace or something else, but don't have a
specific pointer.


- FChE


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]