On 12/1/06, Doug Leavitt <[EMAIL PROTECTED]> wrote:
Hi Joe,
Thanks for joining the e-mail list and bringing this issue to our
attention.
My suspicion is that the core dumps you might have been seeing previously
might have been one of these:
6482827 nscd dumps core when doing getent passwd in compat mode
484895 innetgr.0[012] dumps core on non sparc machines with nscd running
which were both fixed in snv_52. Even though you are no longer getting
core dumps it sounds like this is an issue, that we currently do not
have a bug open on, but that we want to start tracking down immediately.
For the first the first set of sparks putbacks, the new code base now
processes (can process) getents/netgroups requests in nscd when nscd is
active but nscd intentionally is not caching them at the moment.
This was a intentional decision on our part so that we could deliver the
code base sooner, and deliver getent/netgroup caching once we had the
wrinkles in that code. We plan to deliver this additional code as soon
as it is complete.
That aside, the new code base, when nscd is enabled, should be using shared
multiplexed connections to the DS so it should stress the DS less than
pre-sparks.
In the libsldap code where this error message:
libsldap: Status: 7 Mesg: Session error no available conn.
is generated, it is usually the result of the client not being able to
connect to any DS either because they are all down or they have all
refused connections. This should happen more frequently pre-sparks
or sparks without nscd because in those cases connections are not shared.
The fact that you are seeing it in sparks with nscd more than the other
situations points to a serious bug that we need to investigate ASAP.
Can you please send me more details (directly if you wish) so we can
log a high priority bug and chase this down ASAP?
Thanks in advance,
Doug.
Just a reply to the lists, that this bug was worked out between Sun
and myself, and it definitely manifested itself in reproducible
fashion when you have anonymous binds for your ldap connection
pooling.
The BUG will be found via #6500952 when its posted. I do hope that it
makes it into B55, but I'm happy to report that the fix works well and
we no longer have dying NFS services from resource exhaustion on the
name services.
More importantly, I'd like to give my thanks to Doug and co for the
quick turnaround as I started to dump core after core into their
inboxes (pun intended). This is exactly the interaction and rapid
turnaround that will drive OpenSolaris as a community for those
technically capable but not so familiar with the code base.
______________________________________
> sparks-discuss mailing list
> [EMAIL PROTECTED]
> http://opensolaris.org/mailman/listinfo/sparks-discuss
_______________________________________________
opensolaris-discuss mailing list
[email protected]
