T.R | Title | User | Personal Name | Date | Lines |
---|
3263.1 | | VMSNET::P_NUNEZ | | Fri Mar 21 1997 17:45 | 8 |
|
How does PWMON determine system idle time. Again, I've got a customer
running v5.0E who is having performance problems and PWMON is showing
0% idle, yet MON SYSTEM shows the 4-cpu system to be only 135% active.
VAX 6540 running vms 6.1.
Paul
|
3263.2 | single node or cluster ? | KERNEL::BURNST | | Mon Mar 24 1997 04:39 | 7 |
| Have noticed that sometimes this happens if the node in question is not
a cluster and works fine in a clustered environment.
Also more noticble in eairlier versions 5.0e seem to work Ok but the
only systems I tried was a 1 node cluster and a 2 node cluster bother
running 5.0e but I had seen this before.
Trev. UK CSC.
|
3263.3 | | VMSNET::P_NUNEZ | | Mon Mar 24 1997 09:43 | 8 |
|
He saw it on a non-clustered Alpha and on a 4-node VAXcluster that had
only one node (VAX 6540/VMS V6.1) running PATHWORKS at the time.
How is pwmon calculating this value?
Paul
|
3263.4 | a quick incomplete answer | CPEEDY::KENNEDY | Steve Kennedy | Mon Mar 24 1997 11:09 | 9 |
| .3> How is pwmon calculating this value?
It's not - the idle values PWMON shows are calculated by the file
server and retrieved for display by PWMON. Off the top of my head I
don't know the details of how that value is generated within the file
server and unfortunately I can't take the time to look right now. If I
get a few idle minutes to go take a look I'll post my findings.
\steve
|
3263.5 | | VMSNET::P_NUNEZ | | Mon Mar 24 1997 12:13 | 17 |
|
>It's not - the idle values PWMON shows are calculated by the file
>server and retrieved for display by PWMON.
So it's probably using the same code used by the cluster alias load
balancing code:
The 'average_node_idle_percentage' is calculated periodically
based on statistics gathered by VMS' SPISHR services (used by VMS
monitor).
But is there a timeout or something that the "file server" (lmmcp or
lmsrv?) will give up after and display 0%? Possible issue with SPISHR
services (ie, VMS version)?
Paul
|
3263.6 | | UTRTSC::SWEEP | I want a lolly... | Tue Mar 25 1997 03:26 | 5 |
| No timeout, the values are stored in a global memory
area, so they can be retrieved by all processes that
map it.
Adrie
|