dazuko-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Dazuko-devel] Signal handling


From: John Ogness
Subject: Re: [Dazuko-devel] Signal handling
Date: Thu, 28 Aug 2003 22:09:59 +0200
User-agent: Mozilla/5.0 (X11; U; FreeBSD i386; en-US; rv:1.4) Gecko/20030708

Yes,

This is something that should exist. I will add it to the TODO list.

Not just for SIGTERM related events, but also as safety precautions for killed applications. When a SIGKILL is sent, Dazuko in many situations does not receive *any* indication of this. It goes on assigning access control to the non-existant process, causing the machine to have a hanging effect. :(

Dazuko has severe problems when dealing with "unfriendly" situations (SIGSTOP, SIGKILL, exitting without unregistering). This must be improved.

John Ogness


S Wade wrote:
Is it realistic to search the process table for the application after receiving a SIGTERM to determine whether or not to unregister the application?

Swade


From: John Ogness <address@hidden>
To: Jaroslav Suchanek <address@hidden>
CC: Dazuko devel <address@hidden>
Subject: Re: [Dazuko-devel] Signal handling
Date: Thu, 28 Aug 2003 12:22:47 +0200

Hi,

Which version of Dazuko are you using? This is a bit of a controversial
issue, and I am not sure how to handle it.

In the new pre-release version of Dazuko, a SIGTERM to the registered
application will not only cause dazukoGetAccess to return with an error,
but it will also unregister the applicatoin with Dazuko!

This would be dangerous for applications that will continue working,
even though a SIGTERM has been sent. The advantage of this, is it allows
Dazuko to avoid problems of applications not unregistering themselves
before they quit. However, I am not happy about this and will be
changing it before the next version is released.

So what should Dazuko do?

Do you think dazukoGetAccess should continue blocking, even though a
SIGTERM has been sent? In my opinion this is not an option, since the
application *must* have a chance to exit if it wants.

A good solution would be if different error codes were returned (rather
than simply !0). Then you could know if dazukoGetAccess failed because
of some real problem, or because an interruption signal was sent. What
do you think?

John Ogness


Jaroslav Suchanek wrote:
> Hi,
>
> Is it correct, that after sending signal (e.g. SIGTERM) function
> dazukoGetAccess always return error code?
>
> Is there any way, how to handle signal and continue working with dazuko?
>
> Thanks for all ideas,
>
> Jaroslav Suchanek





reply via email to

[Prev in Thread] Current Thread [Next in Thread]