-----BEGIN PGP SIGNED MESSAGE-----
Hash: RIPEMD160

Alan Altmark wrote:
| On Wednesday, 03/29/2006 at 09:31 ZE2, Grega Bremec <[EMAIL PROTECTED]>
| wrote:
|>
|> We've just upgraded from z/VM 4.4 to z/VM 5.2 and I am wondering if
|> it was safe to use QIOASSIST with OSA and HiperSocket devices with
|> 5.2 already. We had some nasty issues with JDBC connection pools
|> from WAS to DB2 - the kernel did not seem to be cleaning them up
|> properly under 4.4, which caused the TCP session to appear to be
|> alive to both sides, when infact it was not so any more.
|
| QIOASSIST has nothing to do with application or subsystem logic.
| Rather, it allows a guest using dedicated subchannels on real OSD
| (OSA QDIO), FCP, or IQD (HiperSocket) chpids to stay in SIE,
| providing a performance boost. (If your machine has QIOASSIST
| microcode on it.)

Thank you, Alan, for taking the time to explain this. However, I am
very well aware of it.

I apologize, maybe I was not clear enough: I was talking about APAR
VM63685, which applies to z/VM 4.4 and 5.1 and I was interested if
anyone was still experiencing that kind of behaviour on z/VM 5.2:

~ QIOAssist on with either a Guest LAN, OSA Express, HiperSockets or a
~ FCP device, it is possible for the QDIO input queue for the device to
~ stall. The problem will happen when there is a substantial amount of
~ conventional I/O occurring for a guest who has a fairly active QDIO
~ connection.

Since the only types of connections exhibiting that kind of behaviour on
our systems were the JDBC connections between WAS and DB2 users (due to
connection pooling which kept connections alive long enough and the fact
that only the input queue had stalled, which prevented the kernel from
discovering it was impossible to send), I used that as an example of
what kind of situations the behaviour I'm looking for could be observed in.

I don't know if this APAR was closed prior to release of z/VM 5.2, and
last time I heard about it, it was still open. So I'm asking if anyone
can either verify it was actually fixed in z/VM 5.2, if there is a PTF
for it available, or the only workaround at VM level is to set
NOQIOASSIST with DEDICATEd devices.

Kind regards,
- --
~    Grega Bremec
~    gregab at p0f dot net
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFEKsrdfu4IwuB3+XoRA6UsAJ4qtnbOilj2ftT9Tywj+f/ia6u+vwCfU/Zy
WTDnfuWpHJp9cr1ZwDz503s=
=/shf
-----END PGP SIGNATURE-----

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to