Re: Oninits Running Amok...

OK - got some more info...it looks like the only time the CPUVP's run amok is when our Powerbuilder client connects across a slow (dialup) IP link and is downloading client code changes from a table (Select...) that contains updated app code in BLOB's. This would explain the lack of buffer activity, since blobs don't pass through shared memory buffers. When I connect over a very fast link (i.e. from my cube over the LAN, FDDI bacbone etc, CPU activity is nearly nonexistant, as I would expect. It would appear that 7.13 perhaps in conjunction with Solaris 2.4 has problems with slow IP links. This seems to apply regardless of comm server (2 different ones), client IP stack (3 to 4 different ones), etc... Had two crashes this afternoon, both users were dialed up at the time - so this may be turning into a nasty one. Any ideas appreciated...gm. David Williams wrote: > snip... > Try running > > - onstat what does the first line say, is the system checkpointing? > - onstat -C several times - is the system page cleaning > - onstat -D sevel times - look at page Rd/page Wr - is the system > accessing the disk? > - onstat -F (is the system writing to disk > - onstat -g iof is the system writing to temporary sort files/ > workfiles (Not this value includes those from onstat -D). > - onstat -g spi Are longspins occuring i.e. too many CPU VPs > configured. > - onstat -g rea - which threads are running. > - onstat -u - what are users sessions doing? > - onstat -g sch are VPs spinning or busy waiting. > > E-mail what answers you get - I'd be interested to know as I've never > heard of oninits using a lot of CPU time wihtout doing disk I/O. > -- > David Williams