I did recover all my data from a backup. Then I did a slapcat without slapd running twice a day, just for paranoia reasons. It did work on the first day for the first two slapcats, but on the second day slapcat hung again. Is my database corrupt again ? What is happening here ? We did use ldbm before, and had never any problems like this. Should I go back to ldbm instead of bdb ?
Marcella
-----Ursprüngliche Nachricht----- Von: Adam Pordzik [mailto:ap@d-dt.de] Gesendet: Thursday, February 15, 2007 3:39 AM An: Voigtländer, Marcella HTK Cc: richton@nbcs.rutgers.edu; openldap-software@openldap.org Betreff: Re: AW: slapcat hangs without error message
Voigtlaender@hansenet.com wrote:
It seems you are right. I did try db_recover, after that slapcat runs, but the data is not complete.
bdb specific.
Furthermore my slapd doesn't start anymore after the recover. Is there anything else I could try to get my data ?
Is everything chown'ed properly after dbrecovery? I assume, slapd runs as another user.
A
--On Friday, February 16, 2007 9:13 PM +0100 Voigtlaender@hansenet.com wrote:
I did recover all my data from a backup. Then I did a slapcat without slapd running twice a day, just for paranoia reasons. It did work on the first day for the first two slapcats, but on the second day slapcat hung again. Is my database corrupt again ? What is happening here ? We did use ldbm before, and had never any problems like this. Should I go back to ldbm instead of bdb ?
No.
It sounds like you must be doing something very wrong, or there's something seriously wrong with your hardware. I slapcat my database every single night, and have for several years. Of course, you don't provide a whole lot of information about your setup, so it is hard to say what could be happening.
--Quanah
-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
Its a slapd 2.2.23 (debian package) with a bdb Backend on a Debian Sarge system. Transaction Logging is enabled, I set the cachesize in DB_CONFIG to 1 GB. I have one Master and two Slaves. On one of the Slaves the slapcat worked until yesterday, today it is broken again. The other Slave is ok. I see no errors in syslogs. My id2entry.bdb is 1,6 GB, in a complete ldif-File I have 31 Mio lines. Could this be a problem ?
-----Ursprüngliche Nachricht----- Von: Quanah Gibson-Mount [mailto:quanah@stanford.edu] Gesendet: Sonntag, 18. Februar 2007 23:44 An: Voigtländer, Marcella HTK; ap@d-dt.de Cc: openldap-software@openldap.org Betreff: Re: AW: AW: slapcat hangs without error message
--On Friday, February 16, 2007 9:13 PM +0100 Voigtlaender@hansenet.com wrote:
I did recover all my data from a backup. Then I did a
slapcat without
slapd running twice a day, just for paranoia reasons. It
did work on
the first day for the first two slapcats, but on the second day slapcat hung again. Is my database corrupt again ? What is
happening
here ? We did use ldbm before, and had never any problems
like this.
Should I go back to ldbm instead of bdb ?
No.
It sounds like you must be doing something very wrong, or there's something seriously wrong with your hardware. I slapcat my database every single night, and have for several years. Of course, you don't provide a whole lot of information about your setup, so it is hard to say what could be happening.
--Quanah
-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
--On Tuesday, February 20, 2007 9:03 AM +0100 Voigtlaender@hansenet.com wrote:
Its a slapd 2.2.23 (debian package) with a bdb Backend on a Debian Sarge system. Transaction Logging is enabled, I set the cachesize in DB_CONFIG to 1 GB. I have one Master and two Slaves. On one of the Slaves the slapcat worked until yesterday, today it is broken again. The other Slave is ok. I see no errors in syslogs. My id2entry.bdb is 1,6 GB, in a complete ldif-File I have 31 Mio lines. Could this be a problem ?
I'd suggest pursuing this with Debian. As a user of debian, I'm sure you are aware that they patch source code with things that are not found in the master source, and those things could always be the potential source of bugs or other types of errors. I would advise not using their package, and instead either getting a prebuilt distribution like CDS from Symas Corporation, or building it yourself.
--Quanah
-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
Quanah Gibson-Mount wrote
--On Tuesday, February 20, 2007 9:03 AM +0100 Voigtlaender@hansenet.com wrote:
Its a slapd 2.2.23 (debian package) with a bdb Backend on a Debian Sarge system. Transaction Logging is enabled, I set the cachesize in DB_CONFIG to 1 GB. I have one Master and two Slaves. On one of the Slaves the slapcat worked until yesterday, today it is broken again. The other Slave is ok. I see no errors in syslogs. My id2entry.bdb is 1,6 GB, in a complete ldif-File I have 31 Mio lines. Could this be a problem ?
I'd suggest pursuing this with Debian. As a user of debian, I'm sure you are aware that they patch source code with things that are not found in the master source, and those things could always be the potential source of bugs or other types of errors. I would advise not using their package, and instead either getting a prebuilt distribution like CDS from Symas Corporation, or building it yourself.
--Quanah
-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
Also a Debian user here, and I can pretty strongly recommend just building it yourself. A ./configure --help will reveal an amazing wealth of options available to you. I tried a few combinations of the packages before going ahead and building it myself, and I think it's really worth going through that process to get the OpenLDAP implementation you want.
As an added bonus, you get all of your ldap applications and libraries to be the same version. Speaking specifically of what we've got in Debian Stable right now, some packages depend on ldap libraries from OpenLDAP 2.1.17, and some depend on libraries built from 2.2.23. This should soon be fixed as packages make the move from unstable/testing into stable, but it's all no big deal if you're at least content running software labeled "unstable".
As a final note...if you do use the debian packages, be very sure to dive into the documents included with the packages (usually /usr/share/<packagename>), as debian tends to put files in unexpected places that eventually make sense but initially confuse the hell out of you.
-- Sean
--On Tuesday, February 20, 2007 1:19 AM -0800 Quanah Gibson-Mount quanah@stanford.edu wrote:
--On Tuesday, February 20, 2007 9:03 AM +0100 Voigtlaender@hansenet.com wrote:
Its a slapd 2.2.23 (debian package) with a bdb Backend on a Debian Sarge system. Transaction Logging is enabled, I set the cachesize in DB_CONFIG to 1 GB. I have one Master and two Slaves. On one of the Slaves the slapcat worked until yesterday, today it is broken again. The other Slave is ok. I see no errors in syslogs. My id2entry.bdb is 1,6 GB, in a complete ldif-File I have 31 Mio lines. Could this be a problem ?
I'd suggest pursuing this with Debian. As a user of debian, I'm sure you are aware that they patch source code with things that are not found in the master source, and those things could always be the potential source of bugs or other types of errors. I would advise not using their package, and instead either getting a prebuilt distribution like CDS from Symas Corporation, or building it yourself.
One other note, OpenLDAP 2.2.23 is *extremely* old. The last release in the 2.2 series was 2.2.30, and the current release is 2.3.34.
--Quanah
-- Quanah Gibson-Mount Principal Software Developer ITS/Shared Application Services Stanford University GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html
On Feb 20, 2007, at 12:03 AM, Voigtlaender@hansenet.com wrote:
Its a slapd 2.2.23 (debian package) with a bdb Backend on a Debian Sarge system. Transaction Logging is enabled, I set the cachesize in DB_CONFIG to 1 GB. I have one Master and two Slaves. On one of the Slaves the slapcat worked until yesterday, today it is broken again. The other Slave is ok. I see no errors in syslogs. My id2entry.bdb is 1,6 GB, in a complete ldif-File I have 31 Mio lines. Could this be a problem ?'
Something like that happened here, a year ago, when a lock got wedged and all access to the locked part of the directory hierarchy would wait forever for it, including slapcat which of course needs access to the entire hierarchy.
db_recover can resolve this situation, of course with slapd out of service. If that works for you, then I suppose you're running into the same problem. I have no suggestion for how to avoid it, though. It's clearly a failure on the server side, so whoever's responsible for the software there needs to look into it. If that's you, and not your platform vendor, then you probably should start with more recent software.
By the way, did you know that the "Re:" commonly seen in email subject lines is of Latin origin? It's quite international.
Donn Cave, donn@u.washington.edu
openldap-software@openldap.org
-
Donn Cave -
Quanah Gibson-Mount -
Sean Myers -
Voigtlaender＠hansenet.com