--On Thursday, January 17, 2019 9:09 AM +0100 Ulrich Windl Ulrich.Windl@rz.uni-regensburg.de wrote:
Quanah Gibson-Mount quanah@symas.com schrieb am 16.01.2019 um 19:03 in
Nachricht <9C84644093A248A4CD6F4653@[192.168.1.39]>:
‑‑On Wednesday, January 16, 2019 5:37 PM +0000 Howard Chu hyc@symas.com wrote:
Must point out ‑ this is not a fatal error, otherwise slapd would not have continued startup. The backend inits a new shared memory region and just starts up as normal, there's no reason to take any special action here.
Hm, back when I used back‑hdb + SHM, I have to say I didn't run into this issue with OpenLDAP unless something had cleared the shared memory region (i.e., it persisted across restarts of slapd).
So while back‑hdb is clearly recovering, there still seems to me to be an issue here in that short of a server reboot, the SHM segment should remain unaffected.
AFAIR, those memory files (e.g. __db.001) were needed especially for recovering BDB, that is when the database wasn't shut down cleanly. There seem to be two problems then:
- Unclean shutdown of the database
- memory files vanishing
I'm talking about SHM shared memory keys. See the slapd-bdb(5)/slapd-hdb(5) man pages for the shm_key option.
--Quanah
--
Quanah Gibson-Mount Product Architect Symas Corporation Packaged, certified, and supported LDAP solutions powered by OpenLDAP: http://www.symas.com