Thanks Willian and David,
The first tunning already helped on making the database more responsive, now i'm going to make the others changes and monitor the environment.
Thanks a lot.
Em qui, 17 de ago de 2017 às 20:50, William Brown <wibrown@xxxxxxxxxx> escreveu:
On Thu, 2017-08-17 at 10:53 +1000, William Brown wrote:
> On Tue, 2017-08-15 at 13:15 +0000, Lucas Diedrich wrote:
> > Willian,
> >
> > Cache values from cn=config + cn=userRoot:
> > https://paste.fedoraproject.org/paste/ZqmA2wUkQDcSUaUIcpGDhg
> > free -h + lspcu:
> > https://paste.fedoraproject.org/paste/Br8Vz5-quxtcatiDyMy3QA
> >
> > I'll check it out the docs for the optimization + wait until your response,
> > i think it will more secure.
>
> Hey mate,
>
> Those values are all default of 10MB. You probably want to change:
>
> dn: cn=userRoot,cn=ldbm database,cn=plugins,cn=config
> nsslapd-cachememsize: 10485760
> nsslapd-dncachememsize: 10485760
>
> and
>
> dn: cn=config,cn=ldbm database,cn=plugins,cn=config
> nsslapd-dbcachesize: 10000000
>
> Looking at your values of free/cpu, I would recommend something like:
>
> dn: cn=userRoot,cn=ldbm database,cn=plugins,cn=config
> nsslapd-cachememsize: 268435456
> nsslapd-dncachememsize: 67108864
>
> and
>
> dn: cn=config,cn=ldbm database,cn=plugins,cn=config
> nsslapd-dbcachesize: 536870912
>
>
> That should help you as a start, and you can tweak these up and down as
> needed.
>
I was forwarded this from another list user too:
Lucas,
You may also want to make additional modifications below in
addition to what William provided.
You currently have transactions building up and until you hit 60
seconds the commits are being held in durability transaction log.
Below setting will keep durability transaction log on, BUT only 1
transaction will be held until 60 seconds
nsslapd-db-transaction-batch-val: 0
to be
nsslapd-db-transaction-batch-val: 1
If you are not doing ldif2db transactions as part of normal
operation have discovered over the
Years that disabling import-cache-autosize helps with memory
management. Historically 389ds will fence this
Memory not use it and never release it back to OS.
nsslapd-import-cache-autosize: -1
nsslapd-import-cachesize: 20000000
to be
nsslapd-import-cache-autosize: 0
nsslapd-import-cachesize: 20000000
Make sure you check cachesize values for change after each directory
restart. They are dynamic based on available memory on OS at time of
service start.
If you have something else running on host with memory leak
Directory Server slowly will be starved and eventually crash.
Memory allocations for each directory instance you have running on
host are cumulative so be aware what else is running on host before
modifying the cachesize settings especially if you leave autosize
enabled.
David M. Partridge
Identity Management and Security Engineer
Tangible Security Inc
2010 Corporate Ridge, Suite 250
McLean, VA 22102
I hope this helps you too,
--
Sincerely,
William Brown
Software Engineer
Red Hat, Australia/Brisbane
_______________________________________________
389-users mailing list -- 389-users@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to 389-users-leave@xxxxxxxxxxxxxxxxxxxxxxx
_______________________________________________ 389-users mailing list -- 389-users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to 389-users-leave@xxxxxxxxxxxxxxxxxxxxxxx
