ldbadd --no-sync implications?

[Michael Wood]

Hi metze

Thanks for your reply.

On 3 March 2011 18:11, Stefan (metze) Metzmacher <metze at samba.org> wrote:
> Hi Michael,
>
>> On 3 March 2011 11:54, Michael Wood <esiotrot at gmail.com> wrote:
>> [...]
>>> What's a good way to add a bunch of users in a reasonably short amount
>>> of time while not causing other operations to be impacted too badly?
>>> Would it help if I set up another Samba box and replicated from one to
>>> the other?
>>
>> By the way, using ldbadd --nosync to add 1000 users took around 1.5
>> minutes.  The strace output was strange...
>
> I'm using TDB_NO_FSYNC=1; export TDB_NO_FSYNC, so that transactions
> doesn't trigger real disk io.

Is this equivalent to ldbadd/modify --nosync?

So using TDB_NO_FSYNC=1, I assume is safe as long as you don't crash
while it's running?

It seems, though, that disk I/O is not really the problem for me.
It's using up 100% of one of my cores, but seems not to do much disk
I/O (I am using the --nosync option.)

> Adding about 15000 user, groups and computers with about 90000
> group membership (batched to ldif modify per group) takes about 10 mins
> for me.

Oh that's good news!  For me at the moment, adding 1000 users to the
group that already has 30000 users in it is taking:

real	9m19.655s
user	9m11.270s
sys	0m7.020s

How are you doing the add and how many users do you have in the largest groups?

Should it work OK with 40000 users all in the same group?

-- 
Michael Wood <esiotrot at gmail.com>