On Fri, Apr 24, 2015 at 07:58:46AM +0800, Jones Syue wrote:> Hello Jeremy, > > Sure here is the link to patch: > https://drive.google.com/file/d/0B7hCXnjbSiXFWDltM2lwcVdCdGs/view?usp=sharing > > Based on this push: > s3:passdb: do not use the in-memory idmap cache (it will be removed) > https://git.samba.org/?p=samba.git;a=commit;h=c89affbd8da230cae6df25558fe621510690392c > > Not quite sure why remove it from git, > perhaps will match new features or something new. > Loop Michael too, > if could help review and comment that will be great help, > thank you.OK, this functionality got moved into idmap_cache_find_XXXX() functions which allow smbd to look directly into the winbindd idmap cache via the uid_to_sid(), gid_to_sid(), sid_to_uid(), sid_to_gid() and sids_to_unixids() functions. These functions explicitly look inside the gencache tdb used for all generic caching (with expiry time) inside Samba, so this should be pretty efficient. Does this extra layer of in-memory caching really make such a difference on top of the caching we are already using in the gencache tdb ? Any chance of instrumenting this without your patch to find out why is makes such a difference ? Michael, do you have any good ideas on why removing the in-memory cache hurt performance so much ? Jeremy.
On Fri, May 29, 2015 at 04:53:07PM -0700, Jeremy Allison wrote:> Michael, do you have any good ideas on why removing > the in-memory cache hurt performance so much ?fcntl locks? Under very heavy load this might make a difference. Volker -- SerNet GmbH, Bahnhofsallee 1b, 37081 G?ttingen phone: +49-551-370000-0, fax: +49-551-370000-9 AG G?ttingen, HRB 2816, GF: Dr. Johannes Loxen http://www.sernet.de, mailto:kontakt at sernet.de
? On Sat, May 30, 2015 at 7:53 AM, Jeremy Allison <jra at samba.org> wrote:> Does this extra layer of in-memory caching really > make such a difference on top of the caching we > are already using in the gencache tdb ? > > Any chance of instrumenting this without your > patch to find out why is makes such a difference ?? On Sat, May 30, 2015 at 5:27 PM, Volker Lendecke <Volker.Lendecke at sernet.de> wrote:> fcntl locks? Under very heavy load this might make a > difference.Hello Jeremy, Volker, Thank for kindly feedback! Sure i would like to dig further but might take a bit. One more thing to mention on my earlier test is using local users instead of domain users, perhaps i should test with domain users, which are more close to Ervin's case. -- Regards, Jones Syue | ??? QNAP Systems, Inc.
Hi Jones, and all, first: thanks for your help. On Fri, Jun 12, 2015 at 04:15:23PM +0800, Jones Syue wrote:> ? > On Sat, May 30, 2015 at 7:53 AM, Jeremy Allison <jra at samba.org> wrote: > > > Does this extra layer of in-memory caching really > > make such a difference on top of the caching we > > are already using in the gencache tdb ? > > > > Any chance of instrumenting this without your > > patch to find out why is makes such a difference ? > > > ? > On Sat, May 30, 2015 at 5:27 PM, Volker Lendecke <Volker.Lendecke at sernet.de> > wrote: > > > fcntl locks? Under very heavy load this might make a > > difference. > > > Hello Jeremy, Volker, > > Thank for kindly feedback! > Sure i would like to dig further but might take a bit. > > One more thing to mention on my earlier test is > using local users instead of domain users, > perhaps i should test with domain users, > which are more close to Ervin's case.yes, the users are "domain users", stored in LDAP - but the computers aren't joined to domain, there are just in WORKGROUP. thanks, Ervin