[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: 2.3.11 syncrepl failing - send_ldap_response: ber write failed

Howard, my apologies if it seems to you that I am complaining - I tried hard not to appear that way. I've been using OpenLDAP since version 1.x (where x is some small number that escapes me), and if I were the complaining type you'd have certainly heard from me before now. :-)

I'm afraid I was also less than clear about my question re: the Subject: - I was wondering if anyone knew "what it is trying to write when it fails", when I actually meant "what it was trying to write TO when it fails".

That is, I agree with you that WHAT the server was trying to write is irrelevant, and that why the connection disappeared is not - but I am still in the dark as to what the connection might be - is it Network? Disk? UNIX Socket? Memory? Do I have a badly tuned kernel? thrashing RAID array? too many threads? too much logging?

I am aware of ITS #4238 being fixed in 2.3.12 - this is what I meant when I was talking about the missing glue entries - I KNOW it's a bug, I have read the bug database and therefore I know how to work around it (expand my filter to include everything up to the top of the tree), I am looking forward to getting rid of that workaround when I do upgrade, and I am specifically NOT complaining about this.

Ralf Haferkamp was good enough to post SuSE RPMs of 2.3.11, but if I want to upgrade beyond that I am going to have to roll my own RPM, and as Quanah puts so well, that's not something you do without some dev servers and some time.

If I do ever find an actual bug that someone hasn't reported already, you can be sure that I will report it, after double checking the bug database, replicating it at least once, capturing some debug output, posting to the "technical discussion" list (i.e. this one) to ensure that it is actually a bug, and posting it to the correct place. Correct me if I am wrong, but that would be the ITS, not here.

Sorry if I stepped on anyone's toes, I am actually on holiday still and I suppose I should be taking more time composing emails to make sure noone gets the wrong idea - I LUV OpenLDAP and this is the first time I have ever failed to fix a problem by reading the FAQs and DOCs and ITS first, and so I assume that I'm doing something wrong, and could use a little help. If you find any of my posts offensive in any way please feel free to reply personally, off the list.

Howard Chu wrote:

Configuration management is always a problem for any sizable software deployment, and everyone will have to choose the approach that works best for them. But it sounds like you're complaining that we fix bugs too often; perhaps you'd be happier using Microsoft AD or some other commercial product with a more glacial response to bugs. To each their own.

re: why ber_write failed - it's not really relevant what the server was trying to write. What matters is that the connection it was writing on died, thus the failure. You need to identify why the connection disappeared.

re: creating missing glue entries - there's code in the consumer to create these entries. If you can demonstrate a case where they're not created correctly, you should file an ITS. Don't sit on your hands when you encounter a bug. (Oh wait, I forgot that you don't like it when bugs get fixed. Never mind.)

Ron Grant wrote:

Thanks, Quanah, I should have added that I am aware that 2.3.11 is all of three months old, and I am aware of 2.3.14, and am hoping to get the time to build an RPM based on it (we try to avoid CVS software on production servers), and I am sure that eventually the folks who run the www.openldap.org website will see fit to declare 2.3.14 the latest stable release instead of 2.3.11, however for now I cannot "just upgrade", and I have looked at the changelog for 2.3.14 and while I see some syncrepl changes, none of them seem to mention fixing a problem that causes slapd to exit ungraciously.

By the time I get 2.3.14 installed....Oops! Look at that! 2.3.15 is out!

I can't always "Get the latest version" because new versions appear too quickly. I know that it's a waste of everyone's time to be debugging an older version (and if "upgrade" is the answer, then upgrade it is), but it's also a waste of time chasing the newest version only to learn that I didn't have a configuration value set correctly - and 98% of errors I find are the result of my misuse of software, not from actual bugs.

I really only need to know if anyone has any idea what it is trying to write when it fails with a "ber write failed"? I've already "worked around" one bug in syncrepl (the one about not creating glue entries) and would be happy to work around this problem while syncrepl stabilizes. I'm MUCH more comfortable using the "stable" release.....

Quanah Gibson-Mount wrote:

--On Wednesday, January 04, 2006 2:35 PM -0800 Ron Grant <rgrant@shuttle.net> wrote:

I have a SuSE 10.0 Master running 2.3.11 (from the RPMS at
ftp://ftp.suse.com/pub/projects/OpenLDAP/2.3/i386/10.0) and a syncrepl
slave on a busy mailserver (also SuSE 10.0/2.3.11), and configured
Postfix/Courier/saslauthd/Amavis to look to the localhost replica for

Other questions aside, I will note that 2.3.11 is a rather old release at this point, and significant fixes to syncrepl have occurred since that time.