Brandon wrote:
> On Fri, Apr 11, 2008 at 12:21 PM, Jürgen Keil <[EMAIL PROTECTED]> wrote:
> > > And another case was an attempt to use "xsltproc(1)" on a
> > > big xml file, this time on an amd64 x2 machine with 4GB of
> > > memory, using zfs, and the xsltproc process had grown to
> > > use > 2GB of memory. Again heavy disk trashing, and I
> > > didn't had the impression that the arc cache did
> > > shrink enough to prevent that thrashing.
> >
> > It looks like this, in top:
>
> prstat can give more accurate info on Solaris, fyi.
Yep, top doesn't work well with 64-bit processes.
But top's summary information includes quite a few
things that are missing in prstat, like swap space
usage, memory totals, total cpu usage, ...
> > load averages: 0.01, 0.02, 0.05 21:11:47
> > 75 processes: 74 sleeping, 1 on cpu
> > CPU states: 99.4% idle, 0.1% user, 0.5% kernel, 0.0% iowait, 0.0% swap
> > Memory: 4031M real, 108M free, 2693M swap in use, 1310M swap free
> >
> > PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
> > 7587 jk 1 60 0 2430M 2306M sleep 0:25 0.24% xsltproc
> Only 124M of the xsltproc process is paged - The rest is resident, but
> there is 2693M of swap in use. It appears that something other than
> your xsltproc is paged out.
Nope, most of the 2693M of "swap in use" is reserved swap
space for the heap data that is allocated by the big xsltproc
process.
I also had a few old / big files in /tmp, but not more than 50mbytes.
> AFAIK, the ARC should not page,
Correct, that's in the kernel's heap.
> so what else is using that memory?
The big xsltproc process, and a few files in /tmp
Plus the X11 server & dtlogin, but no user
logged in into JDS.
> Your ARC + xsltproc are only equal to
> 2754M. There's something else on your system consuming 3862M.
That's not what I see...
I'd say the ARC consumes arc.c 1356 MB
+ arc.arc_meta_used 448 MB (?);
and there is a 2306MB RSS for the xsltproc process.
That's a total of 4110 MB.
And in case the arc meta data is already included
in arc.c:
1356 MB + 2306MB = 3662 MB used.
> Before you ran the process, what did things look like, memory wise?
Before running xsltproc:
% swap -s
total: 105256k bytes allocated + 16804k reserved = 122060k used, 4133532k
available
And while it is running:
% swap -s
total: 2590132k bytes allocated + 16604k reserved = 2606736k used, 1668740k
available
> Also keep in mind that /tmp is an in memory filesystem. Things written
> there will use physical memory, or page out if required.
>
> Also, could you sort the output by memory usage?
> "prstat -s rss" and
> "prstat -s size" will sort by resident size or total
> image size.
% prstat -s size
PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
22642 jk 2430M 2029M sleep 60 0 0:00:22 19% xsltproc/1
22498 root 40M 15M sleep 59 0 0:00:01 0.0% Xorg/1
9 root 20M 1784K sleep 59 0 0:01:23 0.0% svc.configd/15
15634 postgres 19M 1116K sleep 59 0 0:01:06 0.0% postgres/1
15632 postgres 19M 1380K sleep 59 0 0:00:00 0.0% postgres/1
3036 root 13M 748K sleep 59 0 0:00:27 0.0% fmd/17
22535 root 13M 4516K sleep 59 0 0:00:00 0.0% dtgreet/1
7 root 11M 540K sleep 59 0 0:00:29 0.0% svc.startd/12
3108 root 10M 0K sleep 59 0 0:00:00 0.0% smbd/1
4192 root 10M 0K sleep 59 0 0:00:00 0.0% smbd/1
15635 postgres 9636K 952K sleep 59 0 0:00:01 0.0% postgres/1
15636 postgres 8900K 0K sleep 59 0 0:00:00 0.0% postgres/1
22500 root 8504K 1652K sleep 59 0 0:00:00 0.0% dtlogin/1
4207 smmsp 8036K 552K sleep 59 0 0:00:03 0.0% sendmail/1
4238 root 8028K 1140K sleep 59 0 0:00:48 0.0% sendmail/1
3699 root 7776K 1004K sleep 59 0 0:00:00 0.0% dtlogin/1
3307 root 7080K 1572K sleep 59 0 0:03:22 0.0% intrd/1
365 root 6784K 1976K sleep 59 0 0:00:06 0.0% hald/4
146 daemon 6568K 900K sleep 59 0 0:00:00 0.0% kcfd/3
27392 root 6300K 1796K sleep 59 0 0:03:38 0.0% ypserv/1
26803 root 5576K 2608K sleep 59 0 0:00:56 0.0% nscd/35
137 root 5572K 1052K sleep 59 0 0:00:00 0.0% syseventd/15
2859 root 5068K 1544K sleep 59 0 0:00:27 0.0% inetd/4
430 root 4712K 0K sleep 59 0 0:00:00 0.0% hald-addon-netw/1
22645 root 4508K 2856K cpu1 59 0 0:00:00 0.0% prstat/1
3038 root 4420K 0K sleep 59 0 0:00:00 0.0% sshd/1
6764 root 4108K 1176K sleep 59 0 0:00:11 0.0% syslogd/13
541 root 4032K 732K sleep 59 0 0:00:00 0.0% rmvolmgr/1
% prstat -s rss
PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
22642 jk 2430M 2088M sleep 60 0 0:00:22 0.6% xsltproc/1
22498 root 40M 15M sleep 59 0 0:00:01 0.0% Xorg/1
22535 root 13M 4520K sleep 59 0 0:00:00 0.0% dtgreet/1
26803 root 5576K 2964K sleep 59 0 0:00:56 0.0% nscd/35
22651 root 4508K 2784K cpu1 59 0 0:00:00 0.0% prstat/1
3307 root 7080K 2624K sleep 59 0 0:03:22 0.0% intrd/1
9 root 20M 2216K sleep 59 0 0:01:23 0.0% svc.configd/15
22638 root 3336K 2096K sleep 49 0 0:00:00 0.0% tcsh/1
27393 root 3112K 1980K sleep 59 0 0:00:20 0.0% rpc.nisd_resolv/1
365 root 6784K 1976K sleep 59 0 0:00:06 0.0% hald/4
22600 jk 3524K 1920K sleep 59 0 0:00:00 0.0% tcsh/1
22608 jk 3512K 1900K sleep 59 0 0:00:00 0.0% tcsh/1
22634 jk 3400K 1864K sleep 59 0 0:00:00 0.0% tcsh/1
27392 root 6300K 1820K sleep 59 0 0:03:38 0.0% ypserv/1
22500 root 8504K 1652K sleep 59 0 0:00:00 0.0% dtlogin/1
2876 root 3568K 1612K sleep 59 0 0:01:23 0.0% automountd/5
2859 root 5068K 1544K sleep 59 0 0:00:27 0.0% inetd/4
305 root 3996K 1524K sleep 59 0 0:00:44 0.0% devfsadm/7
22607 root 2632K 1476K sleep 59 0 0:00:00 0.0% in.rlogind/1
22599 root 2632K 1464K sleep 59 0 0:00:00 0.0% in.rlogind/1
15632 postgres 19M 1412K sleep 59 0 0:00:00 0.0% postgres/1
7137 root 2716K 1364K sleep 100 - 0:01:44 0.0% xntpd/1
4274 root 3844K 1324K sleep 59 0 0:00:00 0.0% mountd/5
18588 root 2536K 1312K sleep 59 0 0:00:00 0.0% hald-addon-stor/1
579 daemon 3588K 1292K sleep 59 0 0:01:27 0.0% rpcbind/1
6764 root 4108K 1176K sleep 59 0 0:00:11 0.0% syslogd/13
4238 root 8028K 1156K sleep 59 0 0:00:48 0.0% sendmail/1
Btw, when the xsltproc process starts when memory usage on the
machine looks like this, it completes in less than a minute
(this is after several xsltproc runs, zfs arc has released enough
memory so that the big process fits into free memory):
load averages: 0.00, 0.05, 0.06 16:14:03
71 processes: 70 sleeping, 1 on cpu
CPU states: 99.8% idle, 0.0% user, 0.2% kernel, 0.0% iowait, 0.0% swap
Memory: 4031M real, 2589M free, 123M swap in use, 4057M swap free
PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
22993 jk 1 59 0 3788K 1556K cpu/0 0:00 0.02% top
27392 root 1 59 0 6300K 1860K sleep 3:38 0.00% ypserv
22498 root 1 59 0 0K 0K sleep 0:01 0.00% Xorg
22535 root 1 59 0 13M 4520K sleep 0:00 0.00% dtgreet
4276 daemon 2 60 -20 2756K 720K sleep 5:35 0.00% nfsd
3307 root 1 59 0 7080K 2732K sleep 3:22 0.00% intrd
7137 root 1 100 -20 2716K 1364K sleep 1:44 0.00% xntpd
579 daemon 1 59 0 3588K 1296K sleep 1:27 0.00% rpcbind
9 root 15 59 0 20M 2340K sleep 1:23 0.00% svc.configd
2876 root 5 59 0 3568K 1836K sleep 1:23 0.00% automountd
15634 postgres 1 59 0 19M 1152K sleep 1:06 0.00% postgres
26803 root 35 59 0 5576K 3072K sleep 0:56 0.00% nscd
4238 root 1 59 0 8028K 1192K sleep 0:48 0.00% sendmail
305 root 7 59 0 3996K 1524K sleep 0:44 0.00% devfsadm
7 root 12 59 0 11M 592K sleep 0:29 0.00% svc.startd
Now all I have to do is fill the arc cache with
"find /path/to/a/zfs/filesystem -type f -exec grep does_not_exist {} +",
after that findmemory usage changes like this:
load averages: 0.12, 0.20, 0.14 16:19:44
71 processes: 70 sleeping, 1 on cpu
CPU states: 99.7% idle, 0.1% user, 0.2% kernel, 0.0% iowait, 0.0% swap
Memory: 4031M real, 669M free, 123M swap in use, 2137M swap free
PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
22993 jk 1 59 0 3788K 1556K cpu/1 0:00 0.02% top
22498 root 1 59 0 0K 0K sleep 0:01 0.00% Xorg
22535 root 1 59 0 13M 4520K sleep 0:00 0.00% dtgreet
4276 daemon 2 60 -20 2756K 720K sleep 5:35 0.00% nfsd
27392 root 1 59 0 6300K 1860K sleep 3:38 0.00% ypserv
3307 root 1 59 0 7080K 2732K sleep 3:22 0.00% intrd
7137 root 1 100 -20 2716K 1364K sleep 1:44 0.00% xntpd
579 daemon 1 59 0 3588K 1296K sleep 1:27 0.00% rpcbind
9 root 15 59 0 20M 2340K sleep 1:23 0.00% svc.configd
2876 root 5 59 0 3568K 1836K sleep 1:23 0.00% automountd
15634 postgres 1 59 0 19M 1152K sleep 1:06 0.00% postgres
26803 root 35 59 0 5576K 3072K sleep 0:56 0.00% nscd
4238 root 1 59 0 8028K 1192K sleep 0:48 0.00% sendmail
305 root 7 59 0 3996K 1524K sleep 0:44 0.00% devfsadm
7 root 12 59 0 11M 592K sleep 0:29 0.00% svc.startd
... and in this environment xsltproc needs > 15 minutes,
because it doesn't get enough RSS:
load averages: 0.30, 0.23, 0.15 16:20:31
72 processes: 71 sleeping, 1 on cpu
CPU states: % idle, % user, % kernel, % iowait, % swap
Memory: 4031M real, 32M free, 2013M swap in use, 1151M swap free
PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
23076 jk 1 60 0 1894M 1556M sleep 0:17 19.38% xsltproc
23078 jk 1 58 0 3724K 1456K cpu/1 0:00 0.09% top
22498 root 1 60 0 0K 0K sleep 0:01 0.01% Xorg
27392 root 1 60 0 6300K 1600K sleep 3:38 0.01% ypserv
15634 postgres 1 59 0 19M 1116K sleep 1:06 0.00% postgres
4238 root 1 59 0 8028K 1044K sleep 0:48 0.00% sendmail
22535 root 1 59 0 13M 1504K sleep 0:00 0.00% dtgreet
4276 daemon 2 60 -20 2756K 564K sleep 5:35 0.00% nfsd
3307 root 1 59 0 7080K 1512K sleep 3:22 0.00% intrd
7137 root 1 100 -20 2716K 1364K sleep 1:44 0.00% xntpd
579 daemon 1 59 0 3588K 1048K sleep 1:27 0.00% rpcbind
2876 root 5 59 0 3568K 1592K sleep 1:23 0.00% automountd
9 root 15 59 0 20M 624K sleep 1:23 0.00% svc.configd
26803 root 35 59 0 5576K 2096K sleep 0:56 0.00% nscd
305 root 7 59 0 3996K 1000K sleep 0:44 0.00% devfsadm
....
load averages: 0.32, 0.24, 0.16 16:21:00
72 processes: 71 sleeping, 1 on cpu
CPU states: 99.3% idle, 0.1% user, 0.6% kernel, 0.0% iowait, 0.0% swap
Memory: 4031M real, 118M free, 2549M swap in use, 952M swap free
PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
23076 jk 1 60 0 2430M 1821M sleep 0:22 8.97% xsltproc
23078 jk 1 59 0 3788K 1408K cpu/1 0:00 0.03% top
22498 root 1 59 0 0K 0K sleep 0:01 0.01% Xorg
27392 root 1 59 0 6300K 1656K sleep 3:38 0.00% ypserv
3307 root 1 59 0 7080K 2356K sleep 3:22 0.00% intrd
9 root 15 59 0 20M 1056K sleep 1:23 0.00% svc.configd
15634 postgres 1 59 0 19M 1152K sleep 1:06 0.00% postgres
22535 root 1 59 0 13M 1504K sleep 0:00 0.00% dtgreet
4276 daemon 2 60 -20 2756K 564K sleep 5:35 0.00% nfsd
7137 root 1 100 -20 2716K 1364K sleep 1:44 0.00% xntpd
579 daemon 1 59 0 3588K 1104K sleep 1:27 0.00% rpcbind
2876 root 5 59 0 3568K 1512K sleep 1:23 0.00% automountd
26803 root 35 59 0 5576K 2404K sleep 0:56 0.00% nscd
4238 root 1 59 0 8028K 1144K sleep 0:48 0.00% sendmail
305 root 7 59 0 3996K 1000K sleep 0:44 0.00% devfsadm
PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
23076 jk 2430M 1831M sleep 60 0 0:00:22 4.6% xsltproc/1
22498 root 40M 8948K sleep 59 0 0:00:01 0.0% Xorg/1
23080 jk 4556K 2956K cpu0 59 0 0:00:00 0.0% prstat/1
26803 root 5576K 2404K sleep 59 0 0:00:56 0.0% nscd/35
3307 root 7080K 2356K sleep 59 0 0:03:22 0.0% intrd/1
22634 jk 3400K 2064K sleep 49 0 0:00:00 0.0% tcsh/1
27393 root 3112K 1732K sleep 59 0 0:00:20 0.0% rpc.nisd_resolv/1
27392 root 6300K 1656K sleep 59 0 0:03:38 0.0% ypserv/1
2876 root 3568K 1512K sleep 59 0 0:01:23 0.0% automountd/5
22535 root 13M 1508K sleep 59 0 0:00:00 0.0% dtgreet/1
15632 postgres 19M 1416K sleep 59 0 0:00:00 0.0% postgres/1
7137 root 2716K 1364K sleep 100 - 0:01:44 0.0% xntpd/1
22500 root 8504K 1332K sleep 59 0 0:00:00 0.0% dtlogin/1
365 root 6784K 1332K sleep 59 0 0:00:06 0.0% hald/4
2859 root 5068K 1308K sleep 59 0 0:00:27 0.0% inetd/4
4238 root 8028K 1156K sleep 59 0 0:00:48 0.0% sendmail/1
15634 postgres 19M 1152K sleep 59 0 0:01:06 0.0% postgres/1
18588 root 2536K 1140K sleep 59 0 0:00:00 0.0% hald-addon-stor/1
3449 root 2536K 1124K sleep 59 0 0:00:00 0.0% hald-addon-stor/1
579 daemon 3588K 1104K sleep 59 0 0:01:27 0.0% rpcbind/1
4274 root 3844K 1100K sleep 59 0 0:00:00 0.0% mountd/5
6764 root 4108K 1072K sleep 59 0 0:00:11 0.0% syslogd/13
9 root 20M 1056K sleep 59 0 0:01:23 0.0% svc.configd/15
305 root 3996K 1000K sleep 59 0 0:00:44 0.0% devfsadm/7
22632 jk 1512K 980K sleep 59 0 0:00:00 0.0% script/1
15635 postgres 9636K 960K sleep 59 0 0:00:01 0.0% postgres/1
146 daemon 6568K 900K sleep 59 0 0:00:00 0.0% kcfd/3
4531 daemon 2664K 896K sleep 60 -20 0:00:00 0.0% nfs4cbd/2
22633 jk 1528K 892K sleep 59 0 0:00:00 0.0% script/1
601 root 2840K 748K sleep 59 0 0:00:00 0.0% keyserv/3
137 root 5572K 748K sleep 59 0 0:00:00 0.0% syseventd/15
2832 root 3040K 612K sleep 59 0 0:00:00 0.0% cron/1
7 root 11M 588K sleep 59 0 0:00:29 0.0% svc.startd/12
Total: 72 processes, 212 lwps, load averages: 0.25, 0.23, 0.15
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]
