T.R | Title | User | Personal Name | Date | Lines |
---|
574.1 | debug output available? | DECWET::CARRUTHERS | Life gets easier when you realize you can't have everything. | Mon Apr 14 1997 10:14 | 1 |
| Can you post the debug output here for us to review?
|
574.2 | I'll have it in a few days. | TRN02::FRASSINO | | Mon Apr 14 1997 13:49 | 5 |
| Hi,
I'll go to the cusomer site on Wednesday, and I'll take the daemon.log.
Up to now, thanks for your attention.
Pierpa
|
574.3 | | DECWET::FARLEE | Insufficient Virtual um...er.... | Mon Apr 14 1997 14:38 | 11 |
| Pierpa,
I cannot think of any reason why the server should take so long.
It's not supposed to contact the clients during startup.
The output from the -D4 run would be most helpful. In the meantime,
we'll try to figure out what is going on here by inspection.
Your customer is big, but we've seen bigger. One in particular
is pushing 1000 clients...
Kevin
|
574.4 | | SANITY::LEMONS | And we thank you for your support. | Wed Apr 23 1997 08:49 | 9 |
| What is an average startup time? With ~130 clients, we're seeing nsrd
consume ~95% of the cpu on an AlphaServer 2000, and run for ~15 minutes
before spawning the other nsr processes. This is with NetWorker Server
for Digital UNIX V4.2B with patches, by the by.
Our upgrade to the V4.3 field test is imminent; I'll take a look at
this issue with the next release.
tl
|
574.5 | this is being worked via email | DECWET::EVANS | NSR Engineering | Thu Apr 24 1997 18:43 | 1 |
| and is under investigation
|
574.6 | The logs are available !!! | TRN02::FRASSINO | | Fri Apr 25 1997 04:48 | 18 |
| Hi all,
I apologize for my silence, but only today I've the logs available.
I've them on the node i1vvp.trn.dec.com ( 16.192.144.83), available using
anonymous ftp,under pub/NSR_AVIO directory. You will find the nsrd -D4
output and the daemon.log in two different Z files. I
n the daemon.log you will see the 50 minutes startuptime of nsrd before
forking all the nsr stuff.
The nsr server is an Alpha 1000 with 256 MB memory, 730 MB swap area.
Any explanation is really well accepted ( V3.2A was absolutely not so
slow in startup )!!
Pierpa
P.S.
To speak with me offline, send me mails on TRN02::frassino or
frassino@trn ( allin1 ).
I will be on vacation next week, and I'll be back the 5th of may.
|
574.7 | Pierpa!! | DECWET::EVANS | NSR Engineering | Thu May 01 1997 15:37 | 10 |
| well, I'm fetching them now, but you typed in "NSR_AVIO", and it really
was "AVIO_NSR" - growl, growl... :-)
thanks for letting me do "ls" to figure it out!!! :-)
I'll take a gander through the files when the transfer is done, and see
if there is something to help us...
Thank you for posting them!! I've been struggling to construct a set of
clients that would approximate this.
|
574.8 | findings... | DECWET::EVANS | NSR Engineering | Thu May 01 1997 16:11 | 16 |
| date 4/16/97, times start: 15:50, end: 17:02
15:51 to 16:38 spent in LMF checks (NetWorker licenses) == 47 minutes
16:38 to 16:41 spent verifying volumes == 3 minutes
16:41 to 17:02 spent checking index files == 21 minutes
Number of clients: 171 (clients), 1 server.
Notes:
each client was granted a license, yet it also was reported as
"already licensed".
I'm going to see what tests can be run up on this to see why licensing
takes so long. More when there's data to report.
|
574.9 | more data | DECWET::EVANS | NSR Engineering | Thu May 01 1997 18:08 | 4 |
| my V4.3 install ran 19 clients (17 clients + 2 exempt) in 15 seconds,
and had no "already licensed" messages.
still thinking...
|
574.10 | Other infos. | TRN02::FRASSINO | | Wed May 07 1997 09:18 | 15 |
| Bruce,
yesterday I was at the customer site and I investigated about my
problem.
First of all I've to say that you see "already licensed" probably
because any client is defined twice.So this should be normal.
In addition I've seen that the Alphaserver 1000 takes ~10 seconds per
client to check the license, and during the license check the CPU is
96-97% in system time, 3-4% user time and 0% idle.
For this reason I suppose that there is no chance, since the hardware
is working at its top.
Do you agree or there is a "tuning trick" to apply, eventually?
Thanks in advance,
Pierpa
|
574.11 | we are going to run tests (someday) | DECWET::EVANS | NSR Engineering | Thu May 15 1997 11:38 | 1 |
| when I have time... famous line around here lately...
|