Hi all We are implementing samba-ldap to act as an nt pdc and are seeing performance problems. We have a 1ghz, 3gb Ram, 36gb box that is running samba-2.2.5 and openldap-2.0.23 under redhat 7.3 with kernel 2.4.18-3. Clients are all Win2k SP3. All the ldap requests are to the localhost interface. The box is acting as the PDC for the domain, and also sharing diskspace and printers. When we get around 30-40 smbd processes running everything slows to a crawl. we have about 30 slapd processes running, and the total process count for the box is about 130 at this point. here is the output of top at this point: 12:46pm up 2 days, 17:14, 3 users, load average: 20.24, 20.26, 20.51 129 processes: 106 sleeping, 23 running, 0 zombie, 0 stopped CPU states: 36.1% user, 63.8% system, 0.0% nice, 0.0% idle Mem: 3229040K av, 3166372K used, 62668K free, 0K shrd, 148480K buff Swap: 513976K av, 0K used, 513976K free 2758060K cached We would like to support about 100 simultaneous users. We were using mandatory server based profiles, but have discontinued them to try and improve performance. So, my questions are: 1. Is the amount of processes desired unreasonable for the hardware? 2. If so, does anybody have some figures on users supported for a particular hardware configuration? 3. We are seeing (using smbstatus) exclusive oplocks on files that are on read-only (both in the share definition and the filesystem permissions) shares. Should this be happening? Could we use fake_oplocks on the share to improve performance? 4. Has anybody had any luck with mandatory server-based profiles? any help or suggestions would be greatly appreciated. John
On Tue, 2002-09-24 at 15:57, John Coston wrote:> 12:46pm up 2 days, 17:14, 3 users, load average: 20.24, 20.26, 20.51 > 129 processes: 106 sleeping, 23 running, 0 zombie, 0 stopped > CPU states: 36.1% user, 63.8% system, 0.0% nice, 0.0% idle > Mem: 3229040K av, 3166372K used, 62668K free, 0K shrd,i have about 20 users on a 900Mhz machine with 100G of storage (only 384M ram) i don't think it's ever been cpu bound as a result of samba activity. that 63.8% system seems out of whack... what is the disk subsystem? brad
Sorry for the confusion - if I run "iostat" I get this:
[root@foo root]# iostat
Linux 2.4.18-3 (foo) 09/24/2002
avg-cpu: %user %nice %sys %idle
16.79 0.00 26.39 56.82
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 11.80 2.24 184.95 541354 44609586
dev8-1 11.80 1.67 184.95 403090 44609586
if I run iostat 1 it loops every second, so the results (for the first
16 seconds) are:
[root@foo root]# iostat 1 16
Linux 2.4.18-3 (foo) 09/24/2002
avg-cpu: %user %nice %sys %idle
16.79 0.00 26.40 56.80
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 11.80 2.24 184.91 541354 44615082
dev8-1 11.80 1.67 184.91 403090 44615082
avg-cpu: %user %nice %sys %idle
31.00 0.00 69.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
37.00 0.00 63.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
26.00 0.00 74.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
40.95 0.00 59.05 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
32.00 0.00 68.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 4.00 0.00 136.00 0 136
dev8-1 4.00 0.00 136.00 0 136
avg-cpu: %user %nice %sys %idle
40.00 0.00 60.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
38.00 0.00 62.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
42.34 0.00 57.66 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 11.71 0.00 165.77 0 184
dev8-1 11.71 0.00 165.77 0 184
avg-cpu: %user %nice %sys %idle
34.07 0.00 65.93 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
36.00 0.00 64.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 6.00 0.00 104.00 0 104
dev8-1 6.00 0.00 104.00 0 104
avg-cpu: %user %nice %sys %idle
37.00 0.00 63.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
42.00 0.00 58.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
36.00 0.00 64.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
41.00 0.00 59.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 0.00 0.00 0.00 0 0
dev8-1 0.00 0.00 0.00 0 0
avg-cpu: %user %nice %sys %idle
41.00 0.00 59.00 0.00
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
dev8-0 9.00 0.00 128.00 0 128
dev8-1 9.00 0.00 128.00 0 128
Is that enough or should I post more? iostat -V reports sysstat
version 4.0.3
thanks
On Tuesday, September 24, 2002, at 02:30 PM, Martin MOKREJ? wrote:
> So where is the true "iostat 1" output? ;)
>
> --
> Martin Mokrejs <mmokrejs@natur.cuni.cz>, <m.mokrejs@gsf.de>
> PGP5.0i key is at http://www.natur.cuni.cz/~mmokrejs
> MIPS / Institute for Bioinformatics <http://mips.gsf.de>
> GSF - National Research Center for Environment and Health
> Ingolstaedter Landstrasse 1, D-85764 Neuherberg, Germany
> tel.: +49-89-3187 3683 , fax:?+49-89-3187 3585
>
some output from ps wauxf: for smbd, all the processes are around this value: parky 1963 3.5 0.1 7488 3452 ? R 07:37 15:41 \_ smbd -D for ldap, all of the processes are around this value: ldap 6150 0.0 0.1 75548 5068 ? S 14:45 0:00 \_ /usr/sbin/slapd -u ldap we have about 30 of each process running right now. If I understand this correctly, it means that I have 30 ldap processes that are each using 75 MB of virtual memory, which would be 2.25 GB virtual memory for all of them. here is the output of cat /proc/sys/vm/bdflush - this seems to indicate that the vm is flushing every 5 seconds as indicated by vmstat? [root@student0 root]# cat /proc/sys/vm/bdflush 40 0 0 0 500 3000 60 0 0 I'll have to wait for downtime this evening to test with ldap down. On Tuesday, September 24, 2002, at 02:51 PM, Martin MOKREJ? wrote:> On Tue, 24 Sep 2002, John Coston wrote: > >> we are using large caches for the ldap indexes - 5 megs each for 10 >> indexes. We are not running mysql. >> I can remove the caches and resart ldap to see what effect that has. > > > First, run "ps wauxf" and look which applicatiuons have highest > numbers in > "MEM" and "VSZ" and "RSS" columns. Save the output. > > Then, just have running "vmstat 1" in one window and in another window > shutdown ldap server. You should see after a while that the number > under > "cache" has significantly decreased. > > > > -- > Martin Mokrejs <mmokrejs@natur.cuni.cz>, <m.mokrejs@gsf.de> > PGP5.0i key is at http://www.natur.cuni.cz/~mmokrejs > MIPS / Institute for Bioinformatics <http://mips.gsf.de> > GSF - National Research Center for Environment and Health > Ingolstaedter Landstrasse 1, D-85764 Neuherberg, Germany > tel.: +49-89-3187 3683 , fax:?+49-89-3187 3585 > >