tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jason Brittain" <>
Subject Re: stale RELEASE-NOTES commentary
Date Tue, 29 Aug 2006 04:51:02 GMT
On 8/23/06, William A. Rowe, Jr. <> wrote:
> Yoav Shapira wrote:
> > Hi,
> > It's not only in the 4.1.31 release, it's even in the latest stable
> > 5.5.17.  Let's change it in the next releases of 4.x and 5.x, starting
> > with 5.5.18 which should happen any day now.
> >
> > I think the advice was decent (and based on empirical evidence IIRC)
> > when originally issued, but that was more than two years ago, and it's
> > now stale or worse.
> Worse.  That's the trouble with LD_ASSUME_KERNEL, it's a bandaid that
> gets gross and crusty when left on too long :)

I agree that this is a bandaid that isn't real pleasant, and I agree that
should upgrade either to a better working JVM or a newer kernel so they do
not have to deal with this situation.  However, not everyone can do that.
Regardless, setting LD_ASSUME_KERNEL like this is the only known way
for Tomcat to stay running with that kind of set up.  Without it, Tomcat
appears to serve pages fine for a while, and eventually locks up, probably
due to the JVM locking up.

If something isn't said in the release notes about this, users in this
will end up with a locked up Tomcat and they won't know why.

As one data point on whether this is a helpful option or not, I have been
running Tomcat 4.1 on my server for around 4 years now with the option
set, and it has been perfectly stable.  Without the option set, Tomcat locks
up within 1 day, and sometimes within hours.

> When I suggested we point
> to README, I ment their corresponding VM's readme, not ours; let the VM
> vendor lead or mislead our users and not take ownership of VM quirks.

By at least mentioning that there is a potential for lock up, you don't take
any ownership of the bandaid.  Also, by mentioning that in some cases one
can set that particular option with no ill effects, and that the user's
may vary, at least they have the option to try it, instead of having to
their Tomcat at least once a day and wonder why they have to.



  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message