send sleep
Ray DiMarcello
recurring problem where a user initiates a transaction that acquires some
locks, but before it commits, it gets "stuck" in send sleep and as time goes
by, begins to block other spids. The process is essentially frozen, holding
x amount of exclusive and shared locks from the work it just did, but not
doing anything else because it's waiting on trying to send something back to
the client. The client PB app just has an hourglass.
We're running a 2-tier app; 11.5.1, Powerbuilder 5. The clients are running
OpenClient 11.1 out of the box, I believe - which may be a problem in
itself.
Any insights will help ...
RD
tantinolfi
sendinf or recieving information.
The best thing to do is searching which command is passed by the PB5
application before having the hour glass and test it with these commands :
- set showplan on
- set statistics on ( I am not sure of this, set statistics io on )
- set noexec on.
and run the incriminate command.
Bye
======================
Thierry Antinolfi
http://pro.wanado.fr/dbadevil
Ray DiMarcello a écrit dans le message <7latag$ct2$1...@autumn.news.rcn.net>...
Tony Imbierski
server process has made a network request to send some data to the client, and
we're waiting for the network layer to come back and say "ok, your data has been
sent". Thus any problem at the network level could cause the process to get
stuck in this state. However there have been some bugs where the server actually
missed the wakeup call. Best talk to tech support to see if the version you have
may have this problem.
cheers,
tonyi
>
Mariano Corral
> Can anyone offer any insights into the send sleep status? We have a
> recurring problem where a user initiates a transaction that acquires
some
> locks, but before it commits, it gets "stuck" in send sleep and as
time goes
> by, begins to block other spids. The process is essentially frozen,
holding
> x amount of exclusive and shared locks from the work it just did, but
not
> doing anything else because it's waiting on trying to send something
back to
> the client. The client PB app just has an hourglass.
>
> We're running a 2-tier app; 11.5.1, Powerbuilder 5. The clients are
running
> OpenClient 11.1 out of the box, I believe - which may be a problem in
> itself.
I've been trying to find a recent post by John McVicker
which I'm sure explains that case, but I didn't find it.
I'm not a programmer, so it's quite hard for me trying
to repeat its contents. Excuse the possible mistakes.
The point was with requesting a bunch of results and
not processing all completely. While the GUI user is
watching at his/her screenfull of data, the application
behind has results pending which causes the send sleep
status and may leave behind some shared locks.
John mentioned a specific PB trick which forced the
flush of those pending results.
Regards,
Mariano Corral
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Mariano Corral
> I've been trying to find a recent post by John McVicker
> which I'm sure explains that case, but I didn't find it.
Found it !
... but now, where is the "View for Bookmarking" link
on the new Deja.com ? Well, banned from including the
formerly wonderful DejaNews URL, I'll try to send you
a copy by email.
The details of John's post are:
NG: sybase.public.sqlserver.administration
Date: 1999/06/02
Title: Re: Send sleep problem
Author: John McVicker <mcvi...@sybase.com>
Message-ID: <37552A11...@sybase.com>
Woyzeck
> I don't quite understand what you mean by 'ambiguous'. Send sleep is when the
> server process has made a network request to send some data to the client, and
> we're waiting for the network layer to come back and say "ok, your data has been
> sent". Thus any problem at the network level could cause the process to get
> stuck in this state. However there have been some bugs where the server actually
> missed the wakeup call.
I've got another issue about the "send sleep"
Some Stored Procs I've made, sometimes enter a "send sleep". After that
they appear to be hanging...
besides, what has the stored procedure, executed on the server, to do
with the network layer?
I've seen this behavior repeatedly on Sybase 11.5 on NT box,
though on virtually identical database/sp on AIX, it never happens.
Any ideas?
--
Woyzeck
write to: woyzeck "at" cybergal.com
or ICQ:38209547
----------------------------------------------
the reply-to adress is intentionally invalid!!!
----------------------------------------------
Anthony Mandic
>
> Can anyone offer any insights into the send sleep status? We have a
> recurring problem where a user initiates a transaction that acquires some
> locks, but before it commits, it gets "stuck" in send sleep and as time goes
> by, begins to block other spids. The process is essentially frozen, holding
> x amount of exclusive and shared locks from the work it just did, but not
> doing anything else because it's waiting on trying to send something back to
> the client. The client PB app just has an hourglass.
>
> We're running a 2-tier app; 11.5.1, Powerbuilder 5. The clients are running
> OpenClient 11.1 out of the box, I believe - which may be a problem in
> itself.
There could be any number of reasons. The client could have issued
a request that does a table scan or a sort. If both of these are
on large data sets, it will wait. Another is as Mariano C. pointed
out. The client has requested a lot of data but is only reading a
portion of it at a time. You could try Sybase auditing to see what
queries the server is seeing, but I think the real solution would be
in debugging the PowerBuilder code.
OC 11.1 has a few minor bugs, but these shouldn't be affecting you
with normal use. You could try getting 11.1.1 with the latest EBF
anyway. You may also want to check your server EBF level.
-am
Tony Imbierski
Woyzeck wrote:
>
> I've got another issue about the "send sleep"
> Some Stored Procs I've made, sometimes enter a "send sleep". After that
> they appear to be hanging...
> besides, what has the stored procedure, executed on the server, to do
> with the network layer?
>
We have had some bugs on NT where the network layer already responded but we missed
the callback. They occurred mostly on SMP systems, especially when load was heavy.
Make sure you are on the very latest ESD. If you can reproduce this on the latest ESD
by simply running a stored proc, raise a case through tech support and put my name on
it. Providing the SP is simple enough for me to set up and reproduce myself, I will be
able to fix it. The difficulty with these problems is getting them to happen under
controlled conditions.
>
> I've seen this behavior repeatedly on Sybase 11.5 on NT box,
> though on virtually identical database/sp on AIX, it never happens.
> Any ideas?
The network architecture we use on AIX and NT are different, so it doesn't surprise me
if one has this problem but the other doesn't. In fact this reinforces the idea that
you are hitting a bug in the ASE/NT network layers. As I said above, check your ESD
level and consider upgrading to 11.5.1.1, as we have fixed some problems in this area.
Failing that, raise a support case.
cheers,
tonyi (ASE/NT sustaining engineer)
Ray DiMarcello
Anthony Mandic wrote in message <3779E0A3.1B9BC9E4@_.agd.nsw.gov.au>...
> There could be any number of reasons. The client could have issued
> a request that does a table scan or a sort. If both of these are
> on large data sets, it will wait.
Yeah, but not in send sleep ...
RD
Anthony Mandic
> Yeah, but not in send sleep ...
I see it when doing a large select and the client hasn't
finished receiving all the data. The server marks the spid's
status as being in a send sleep state.
-am