Hi Javi & Mark
looks better, got RSS from 273 MByte down to 53 mbyte after I disabled
your list of services (only need ssh, nfs + iscsi)
before:
ZONEID NPROC SWAP RSS MEMORY TIME CPU ZONE
0 40 190M 273M 27% 0:01:25 1.5% global
after:
ZONEID NPROC SWAP RSS MEMORY TIME CPU ZONE
0 23 44M 52M 5.1% 0:00:19 0.1% global
-bash-3.00# source freemem.sh -l
\nFree Memory Statistics (solaris10)
----------------------\n
Physical Memory: 1015 MB
Lots Free: 15 MB
Desired Free: 7 MB
Min Free: 3 MB
Free RAM Memory: 500 MB
ZFS ARC size: 171 MB
ZFS ARC min: 64 MB
Free memory available for applications: 607 MB
feedback: \n not escaped properly on the title ;) but looks good. is
the better right now?
I do not run services
Am 29.10.2009 um 09:53 schrieb Javier Conde:
>
> Hi Silvan,
>
> Start stoping all inused services to reduce the memory footprint.
> See details in the following blogs entry:
>
> http://blogs.sun.com/javi/
>
> Then, you can use the following script to see how your memory is
> being used. It will be part of my next blog entry, but I need time
> to finish it. Any comment about the script is welcomed!
>
> Cheers,
>
> Javi
>
> jco at atlas:~$ ./freemem -l
>
> Free Memory Statistics (atlas)
> ----------------------
>
> Physical Memory: 96553 MB
> Lots Free: 1506 MB
> Desired Free: 753 MB
> Min Free: 376 MB
> Free RAM Memory: 16029 MB
>
> ZFS ARC size: 16384 MB
> ZFS ARC min: 3017 MB
>
> Free memory available for applications: 29395 MB
>
> jco at mafalda:~$ ./freemem -l
>
> Free Memory Statistics (mafalda)
> ----------------------
>
> Physical Memory: 3319 MB
> Lots Free: 51 MB
> Desired Free: 25 MB
> Min Free: 12 MB
> Free RAM Memory: 1434 MB
>
> ZFS ARC size: 622 MB
> ZFS ARC min: 311 MB
>
> Free memory available for applications: 1746 MB
>
>
>
>
>
> Silvan Gebhardt wrote:
>> oh cr**
>>
>>
>> anybody on this ml confirming need to upgrade memory first then?
>> Shute. it's the lowest server in a couple of railless systems *grrr*
>>
>>
>> Memory: 1023M phys mem, 332M free mem, 512M total swap, 512M free
>> swap
>>
>>
>>
>>
>> I think this is not enough then? is there a way to trick? I do not
>> want to have a fullcrash when doing sth, I'd rather have it going a
>> bit slower when snapshotting. this can't be it? I have only 3
>> clients on this server
>>
>> so what should I do from this thread? I don't want to wreck my box :(
>>
>>
>> Am 29.10.2009 um 09:36 schrieb Mark Furner:
>>
>>> Hi Silvan
>>>
>>> I wonder if this is a 32-bit thing. I googled using the error
>>> message and came across this thread:
>>>
>>> http://opensolaris.org/jive/thread.jspa?messageID=42261
>>>
>>> HTH
>>>
>>> Regards
>>>
>>> Mark
>>>
>>> hi javi
>>>
>>>
>>> okay, this is weird, considered that I just updated today. but
>>> perhabs
>>> snapshotting needs too many threads ;)
>>>
>>>
>>> to the UDP Thing:
>>> please explain me this:
>>>>> Oct 28 19:07:51 solaris10 /usr/lib/nfs/nfsd[9213]: [ID 786952
>>>>> daemon.error] unexpected TLI event (0x20) on connection-oriented
>>>>> transport(tcp,15):Out of stream resources
>>>>> tell me what the last one means.
>>>
>>>
>>> this sounds like TCP to me? the NFS Clients are all ESXi 4.0
>>>
>>> I had the servers on 256 before the crash, on this exact settings
>>> recommended by Thomas H. (Supposedly on this list, too - he was the
>>> "example" SSE Client recently ;) )
>>>
>>> so got any tip to me? I can give you ssh to the box if that helps ;)
>>>
>>> lg
>>> silvan
>>>
>>>
>>>
>>>
>>> Am 28.10.2009 um 22:45 schrieb Javi:
>>>
>>>> No crash then??
>>>>
>>>> To be sure that kernel core dumps can be saved correctly, check the
>>>> output of the "dumpadm" command without any argument.
>>>>
>>>> Regarding the NFS messages, I supposed this server is a NFS server,
>>>> and it seems you are getting out of connections. You'll need to
>>>> increase the default limits of NFS in the /etc/deault/nfs file:
>>>>
>>>> NFSD_SERVERS=256
>>>> LOCKD_SERVERS=64
>>>>
>>>> And then you to stop/start the NFS server to get it activated.
>>>>
>>>> For the e1000g1 error message, I've found a bug with a reference to
>>>> this, but not reproducible.
>>>> http://bugs.opensolaris.org/view_bug.do?bug_id=6533591
>>>>
>>>> In the bug, it is written:
>>>>
>>>> Encountered this on a T2000 client with the SUT running a harmonic
>>>> inc. benchmark
>>>> (heavy multicast and unicast udp packets streams).
>>>> Message appeared after about four days of the test running.
>>>> Box was hung afterwards, couldn't ping or login via console.
>>>> Tried syncing
>>>> to get a panic dump which failed (panic dump timeout).
>>>>
>>>> In your case, NFS is using UDP, which may explain this.
>>>>
>>>> Javi
>>>>
>>>>
>>>> On Wed, Oct 28, 2009 at 8:57 PM, Silvan Gebhardt <pcdog at pcdog.ch>
>>>> wrote:
>>>>>
>>>>> interesting:
>>>>> /var/adm/messages:
>>>>>
>>>>>
>>>>> Oct 28 15:45:20 solaris10 sshd[29944]: [ID 800047 auth.crit]
>>>>> fatal:
>>>>> Timeout before authentication for 10.2.252.252
>>>>> Oct 28 16:38:18 solaris10 sshd[29955]: [ID 800047 auth.crit]
>>>>> fatal:
>>>>> Timeout before authentication for 10.2.252.252
>>>>> Oct 28 19:07:51 solaris10 /usr/lib/nfs/nfsd[9213]: [ID 786952
>>>>> daemon.error] unexpected TLI event (0x20) on connection-oriented
>>>>> transport(tcp,15):Out of stream resources
>>>>> tell me what the last one means.
>>>>>
>>>>> do I need to adjust more on /etc/default/nfs?
>>>>>
>>>>> /var/crash is empty
>>>>> iPhone photo:
>>>>>
>>>>> e1000g1: DL_UNITDATA_REQ failed: DL_BADDADDR (it's the storage
>>>>> interface)
>>>>>
>>>>>
>>>>>
>>>>> got any clue?
>>>>>
>>>>> thanks
>>>>> silvan
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Am 28.10.2009 um 20:08 schrieb Javi:
>>>>>
>>>>> /var/adm/messages
>>>>>
>>>>> PCdogIT
>>>>> Silvan M. Gebhardt
>>>>> pcdog at pcdog.ch
>>>>>
>>>>>
>>>>>
>>>>> +41 32 512 69 58 VOIP CH
>>>>> +41 52 720 39 88 Analog CH
>>>>> +41 78 810 39 88 Cell CH
>>>>> +1 (505) 480-9905 Cell US
>>>>> +1 (360) 718-3416 Landline US
>>>>> + 1 866-941-3869 Toll Free US
>>>>> +31 85 888 0256 NL
>>>>> +44 1225 63 7104 UK
>>>>>
>>>>>
>>>
>>>
>>> PCdogIT
>>> Silvan M. Gebhardt
>>> pcdog at pcdog.ch
>>>
>>> +41 32 512 69 58 VOIP CH
>>> +41 52 720 39 88 Analog CH
>>> +41 78 810 39 88 Cell CH
>>> +1 (505) 480-9905 Cell US
>>> +1 (360) 718-3416 Landline US
>>> + 1 866-941-3869 Toll Free US
>>> +31 85 888 0256 NL
>>> +44 1225 63 7104 UK
>>>
>>>
>>>
>>> _______________________________________________
>>> ug-chosug mailing list
>>> ug-chosug at opensolaris.org
>>> http://mail.opensolaris.org/mailman/listinfo/ug-chosug
>>
>>
>> PCdogIT
>> Silvan M. Gebhardt
>> pcdog at pcdog.ch
>>
>> +41 32 512 69 58 VOIP CH
>> +41 52 720 39 88 Analog CH
>> +41 78 810 39 88 Cell CH
>> +1 (505) 480-9905 Cell US
>> +1 (360) 718-3416 Landline US
>> + 1 866-941-3869 Toll Free US
>> +31 85 888 0256 NL
>> +44 1225 63 7104 UK
>>
>>
>>
>> _______________________________________________
>> ug-chosug mailing list
>> ug-chosug at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/ug-chosug
>
> #!/usr/bin/ksh
> #
> # javier.f.conde at gmail.com
> # freemem v0.2
> #
>
> LONG=0
> while getopts lh: name
> do
> case $name in
> l) LONG=1;;
> h) /usr/bin/echo "Usage: $0 [-l] [-h]"
> exit 2
> ;;
> esac
> done
>
> host=`/usr/bin/hostname`
> freepages=`/usr/bin/kstat -p unix:0:system_pages:freemem | cut -f 2`
> arcsize=`/usr/bin/kstat -p zfs:0:arcstats:size | cut -f 2`
> arcmin=`/usr/bin/kstat -p zfs:0:arcstats:c_min | cut -f 2`
> size=`/usr/bin/pagesize`
> ((freeramb=$freepages * $size))
> ((freememb=$arcsize + $freeramb - $arcmin))
> ((freeram=$freeramb/1024/1024))
> ((freemem=$freememb/1024/1024))
> ((arc=$arcsize/1024/1024))
> ((arcm=$arcmin/1024/1024))
>
> if [ $LONG -ne 1 ] ; then
> printf "%-15s: RAM: %6d MB ARC: %6d MB Total Free: %6d MB\n" $host
> $freeram $arc $freemem
> else
> (( physm=`/usr/bin/kstat -p unix:0:system_pages:physmem | cut -f 2`
> * $size / 1024 / 1024))
> (( lots=`/usr/bin/kstat -p unix:0:system_pages:lotsfree | cut -f 2`
> * $size / 1024 / 1024))
> (( desf=`/usr/bin/kstat -p unix:0:system_pages:desfree | cut -f 2`
> * $size / 1024 / 1024))
> (( minf=`/usr/bin/kstat -p unix:0:system_pages:minfree | cut -f 2`
> * $size / 1024 / 1024))
> echo "\nFree Memory Statistics ($host)"
> echo "----------------------\n"
> printf "Physical Memory: %6d MB\n" $physm
> printf "Lots Free: %6d MB\n" $lots
> printf "Desired Free: %6d MB\n" $desf
> printf "Min Free: %6d MB\n" $minf
> printf "Free RAM Memory: %6d MB\n\n" $freeram
> printf "ZFS ARC size: %6d MB\n" $arc
> printf "ZFS ARC min: %6d MB\n\n" $arcm
> printf "Free memory available for applications: %d MB\n\n" $freemem
> fi
>
PCdogIT
Silvan M. Gebhardt
pcdog at pcdog.ch
+41 32 512 69 58 VOIP CH
+41 52 720 39 88 Analog CH
+41 78 810 39 88 Cell CH
+1 (505) 480-9905 Cell US
+1 (360) 718-3416 Landline US
+ 1 866-941-3869 Toll Free US
+31 85 888 0256 NL
+44 1225 63 7104 UK
