Some Apache children still gracefully restarting after 8 hours

02/06/2006, 00h54

I invoked a "graceful restart" on our main Apache web service (2.0.46).
About 8 hours later I looked at the mod_status status page and noticed
about 50 (out of 150) child servers were still marked "G". This seems
odd. Accoding to the Apache documentation upon a graceful restart the
parent "advises" the children to exit after finishing with the current
connection. How is it possible that the child would still be connected
after 8 hours? I checked and these non-exiting children aren't doing
any data transfer or processing. They just seem to be stuck.

So, is this normal behavior? If not, is there a way to fix it?

02/06/2006, 16h44

google@macrotex.net wrote:
> I invoked a "graceful restart" on our main Apache web service (2.0.46).
> About 8 hours later I looked at the mod_status status page and noticed
> about 50 (out of 150) child servers were still marked "G". This seems
> odd. Accoding to the Apache documentation upon a graceful restart the
> parent "advises" the children to exit after finishing with the current
> connection. How is it possible that the child would still be connected
> after 8 hours? I checked and these non-exiting children aren't doing
> any data transfer or processing. They just seem to be stuck.
>
> So, is this normal behavior? If not, is there a way to fix it?
>

I've seen this also happen in 2.0.55. At least in one case it was some
problem with solaris 2.8 and IPv6. Since we're not using IPv6 I had to
change:

Listen 80

to

Listen XXX.XXX.XXX.XXX:80

which seemed to cause the problem to go away during graceful restarts.

Jeff Long

02/06/2006, 16h44

google@macrotex.net wrote:
> I invoked a "graceful restart" on our main Apache web service (2.0.46).
> About 8 hours later I looked at the mod_status status page and noticed
> about 50 (out of 150) child servers were still marked "G". This seems
> odd. Accoding to the Apache documentation upon a graceful restart the
> parent "advises" the children to exit after finishing with the current
> connection. How is it possible that the child would still be connected
> after 8 hours? I checked and these non-exiting children aren't doing
> any data transfer or processing. They just seem to be stuck.
>
> So, is this normal behavior? If not, is there a way to fix it?
>

I've seen this also happen in 2.0.55. At least in one case it was some
problem with solaris 2.8 and IPv6. Since we're not using IPv6 I had to
change:

Listen 80

to

Listen XXX.XXX.XXX.XXX:80

which seemed to cause the problem to go away during graceful restarts.

Jeff Long

02/06/2006, 00h54	#1
google@macrotex.net Messages: n/a Hébergeur:	Some Apache children still gracefully restarting after 8 hours I invoked a "graceful restart" on our main Apache web service (2.0.46). About 8 hours later I looked at the mod_status status page and noticed about 50 (out of 150) child servers were still marked "G". This seems odd. Accoding to the Apache documentation upon a graceful restart the parent "advises" the children to exit after finishing with the current connection. How is it possible that the child would still be connected after 8 hours? I checked and these non-exiting children aren't doing any data transfer or processing. They just seem to be stuck. So, is this normal behavior? If not, is there a way to fix it?

02/06/2006, 16h44	#2
Jeff Long Messages: n/a Hébergeur:	Re: Some Apache children still gracefully restarting after 8 hours google@macrotex.net wrote: > I invoked a "graceful restart" on our main Apache web service (2.0.46). > About 8 hours later I looked at the mod_status status page and noticed > about 50 (out of 150) child servers were still marked "G". This seems > odd. Accoding to the Apache documentation upon a graceful restart the > parent "advises" the children to exit after finishing with the current > connection. How is it possible that the child would still be connected > after 8 hours? I checked and these non-exiting children aren't doing > any data transfer or processing. They just seem to be stuck. > > So, is this normal behavior? If not, is there a way to fix it? > I've seen this also happen in 2.0.55. At least in one case it was some problem with solaris 2.8 and IPv6. Since we're not using IPv6 I had to change: Listen 80 to Listen XXX.XXX.XXX.XXX:80 which seemed to cause the problem to go away during graceful restarts. Jeff Long

02/06/2006, 16h44	#3
Jeff Long Messages: n/a Hébergeur:	Re: Some Apache children still gracefully restarting after 8 hours google@macrotex.net wrote: > I invoked a "graceful restart" on our main Apache web service (2.0.46). > About 8 hours later I looked at the mod_status status page and noticed > about 50 (out of 150) child servers were still marked "G". This seems > odd. Accoding to the Apache documentation upon a graceful restart the > parent "advises" the children to exit after finishing with the current > connection. How is it possible that the child would still be connected > after 8 hours? I checked and these non-exiting children aren't doing > any data transfer or processing. They just seem to be stuck. > > So, is this normal behavior? If not, is there a way to fix it? > I've seen this also happen in 2.0.55. At least in one case it was some problem with solaris 2.8 and IPv6. Since we're not using IPv6 I had to change: Listen 80 to Listen XXX.XXX.XXX.XXX:80 which seemed to cause the problem to go away during graceful restarts. Jeff Long

Outils de la discussion
Afficher une version imprimable Envoyer un lien vers cette page par email