Worker Process recycling

[cmay]

There are lots of reasons that the WPs can be recycled or killed.

When a WP is recycled, all the InProc sesssion data is lost correct?
Is there any way to a) find out WHY the recycle happened e.g. failed
requests, memory leak, scheduled recycle, lack of requests etc b)
programmatically get a notification that a recycle is going to occurr
to notify the user that their session data is about to be lost,
allowing them to save their work.

Thanks! Thats if for today

 

[Scott Allen]

Generally the machine's event log will contain information about why
the process shut down.

It's impossible to reach the client machine from the server machine to
warn the user that the worker process is going to terminate. The
server and client remain disconnected 99% of the time.

 

[cmay]

Excellent. We had so much junk in the eventlog (a program that should
have been uninstalled...) I didn't even notice.

I have already built a framework for notifying users of pending changes
to the server. Our pages have script that call back to the server
every 5 minutes, checking for messages. So, for example, if we need to
down a server for whatever reason, we can notify everyone who is using
the server within 5 minutes. So we just tell them the server is going
to be taking offline at a specific time and we don't get all the calls
from people who filled out a ton of into on some form only for the
submission to fail.

I was thinking if I could get a warning that the WP was about to
recycle, I could warn the user in this manner. Alternatively, I think
if I tried to schedule the WP to recycle every night at like 3AM, I
could just have a standing rule to tell anyone logged into the server
at 2:50 AM that they should save their data.

 

[Scott Allen]

Excellent. We had so much junk in the eventlog (a program that should
have been uninstalled...) I didn't even notice.

Glad to be of help.

I have already built a framework for notifying users of pending changes
to the server. Our pages have script that call back to the server
every 5 minutes, checking for messages. So, for example, if we need to
down a server for whatever reason, we can notify everyone who is using
the server within 5 minutes. So we just tell them the server is going
to be taking offline at a specific time and we don't get all the calls
from people who filled out a ton of into on some form only for the
submission to fail.

That sounds very cool.

I was thinking if I could get a warning that the WP was about to
recycle, I could warn the user in this manner. Alternatively, I think
if I tried to schedule the WP to recycle every night at like 3AM, I
could just have a standing rule to tell anyone logged into the server
at 2:50 AM that they should save their data.

Unfortunately I think there is very little time between the
Application_End event and the actual recycle that it might not be time
to notify users properly.