all, Thnks to everyone for support. Arup
On Thu, Apr 29, 2010 at 5:30 PM, <[email protected]>wrote: > Send ldoms-discuss mailing list submissions to > [email protected] > > To subscribe or unsubscribe via the World Wide Web, visit > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > or, via email, send a message with subject or body 'help' to > [email protected] > > You can reach the person managing the list at > [email protected] > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of ldoms-discuss digest..." > > > Today's Topics: > > 1. High tcp_listendrop (Tony MacDoodle) > 2. link aggragation in virtual switch (Mike DeMarco) > 3. Performance Bad on Virtual Disks (Tony MacDoodle) > 4. Re: Performance Bad on Virtual Disks (John) > 5. Re: Performance Bad on Virtual Disks (Tony MacDoodle) > 6. Re: High tcp_listendrop (Nathan Kroenert) > 7. Re: Performance Bad on Virtual Disks (Nathan Kroenert) > 8. Re: Performance Bad on Virtual Disks (John) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Wed, 28 Apr 2010 10:17:17 -0400 > From: Tony MacDoodle <[email protected]> > To: [email protected] > Subject: [ldoms-discuss] High tcp_listendrop > Message-ID: > <[email protected]> > Content-Type: text/plain; charset="iso-8859-1" > > Hello, > > Would network saturation cause the following: > > tcp_listendrop * 625597* > * > * > ------ Utilisation ------ ------ Saturation ------ > Date Time %CPU %Mem %Disk %Net CPU Mem > Disk Net > Tue Apr 27 17:22:11 2010 38.45 18.28 0.00 191.12 1.18 0.00 > 0.00 0.00 > Tue Apr 27 17:22:16 2010 2.72 18.29 0.00 19.13 0.20 0.00 > 0.00 0.00 > Tue Apr 27 17:22:21 2010 21.98 18.50 0.00 98.57 0.39 0.00 > 0.00 0.00 > * > * > * > * > * > * > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: < > http://mail.opensolaris.org/pipermail/ldoms-discuss/attachments/20100428/4bfa831e/attachment-0001.html > > > > ------------------------------ > > Message: 2 > Date: Wed, 28 Apr 2010 07:48:47 PDT > From: Mike DeMarco <[email protected]> > To: [email protected] > Subject: [ldoms-discuss] link aggragation in virtual switch > Message-ID: <326903865.121272466157197.javamail.tweb...@sf-app1> > Content-Type: text/plain; charset=UTF-8 > > Anyone see any compelling reason why not to use link aggregation in a > virtaul switch. > > My tests show it works very well as a substitute for ipmp and allows the > bandwidth of several nics. > > TIA > -- > This message posted from opensolaris.org > > > ------------------------------ > > Message: 3 > Date: Wed, 28 Apr 2010 11:10:24 -0400 > From: Tony MacDoodle <[email protected]> > To: [email protected] > Subject: [ldoms-discuss] Performance Bad on Virtual Disks > Message-ID: > <[email protected]> > Content-Type: text/plain; charset="iso-8859-1" > > We have guest domains using ZFS volume groups (RAID 10) that are > experiencing very bad performance. Can I do some more in-depth analysis > using kstat? > LDoms 1.3.... > > Thanks > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: < > http://mail.opensolaris.org/pipermail/ldoms-discuss/attachments/20100428/6a9d60a8/attachment-0001.html > > > > ------------------------------ > > Message: 4 > Date: Wed, 28 Apr 2010 15:24:09 PDT > From: John <[email protected]> > To: [email protected] > Subject: Re: [ldoms-discuss] Performance Bad on Virtual Disks > Message-ID: <250025242.391272493489198.javamail.tweb...@sf-app1> > Content-Type: text/plain; charset=UTF-8 > > We're seeing similar performance problems as well. What are the details of > your configuration, ie what kind of disk's, what storage array, etc > -- > This message posted from opensolaris.org > > > ------------------------------ > > Message: 5 > Date: Wed, 28 Apr 2010 19:35:48 -0400 > From: Tony MacDoodle <[email protected]> > To: John <[email protected]> > Cc: [email protected] > Subject: Re: [ldoms-discuss] Performance Bad on Virtual Disks > Message-ID: > <[email protected]> > Content-Type: text/plain; charset="iso-8859-1" > > T5140's, LDom 1.3, Solaris 10 u8, ZFS for both backend (RAID 10, internal > 146G 10K rpm) and rpool...... Horrible performance for a web server.... We > are thinking on not using the product............ > > On Wed, Apr 28, 2010 at 6:24 PM, John <[email protected]> wrote: > > > We're seeing similar performance problems as well. What are the details > of > > your configuration, ie what kind of disk's, what storage array, etc > > -- > > This message posted from opensolaris.org > > _______________________________________________ > > ldoms-discuss mailing list > > [email protected] > > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > > > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: < > http://mail.opensolaris.org/pipermail/ldoms-discuss/attachments/20100428/d3869a54/attachment-0001.html > > > > ------------------------------ > > Message: 6 > Date: Thu, 29 Apr 2010 10:16:20 +1000 > From: Nathan Kroenert <[email protected]> > To: Tony MacDoodle <[email protected]> > Cc: [email protected] > Subject: Re: [ldoms-discuss] High tcp_listendrop > Message-ID: <[email protected]> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Not specifically network saturation - but something a little more > specific... > > Take a look at > > http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/common/inet/tcp/tcp_input.c > > around line 1362. > > 1362 if (listener->tcp_conn_req_cnt_q >= > listener->tcp_conn_req_max) { > 1363 mutex_exit(&listener->tcp_eager_lock); > 1364 TCP_STAT(tcps, tcp_listendrop); > > > Seems like each time this is true: > listener->tcp_conn_req_cnt_q >= listener->tcp_conn_req_max > > we will increment the counter. > > I'm expecting this is more about the rate of connection requests to the > system rather than non-specific network saturation (which could manifest > in a variety of ways.). > > This blog seems to have some good detail on it: > http://blogs.sun.com/terrygardner/entry/solaris_tcp_ip_parameters_tcp > > As does the Solaris tunables guide... > > > http://docsun.cites.uiuc.edu/sun_docs/C/solaris_9/SUNWaadm/SOLTUNEPARAMREF/p34.html > > and ndd -get /dev/tcp can show you the current values. > > What problem are you trying to solve? Are you simply trying to > understand the metric, or is there an underlying issue you are > experiencing that's caused you to pick out this metric? > > Cheers! > > Nathan. > > Tony MacDoodle wrote: > > Hello, > > > > Would network saturation cause the following: > > > > tcp_listendrop * 625597* > > * > > * > > ------ Utilisation ------ ------ Saturation ------ > > Date Time %CPU %Mem %Disk %Net CPU Mem > > Disk Net > > Tue Apr 27 17:22:11 2010 38.45 18.28 0.00 191.12 1.18 0.00 > > 0.00 0.00 > > Tue Apr 27 17:22:16 2010 2.72 18.29 0.00 19.13 0.20 0.00 > > 0.00 0.00 > > Tue Apr 27 17:22:21 2010 21.98 18.50 0.00 98.57 0.39 0.00 > > 0.00 0.00 > > * > > * > > * > > * > > * > > * > > > > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > ldoms-discuss mailing list > > [email protected] > > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > > > > ------------------------------ > > Message: 7 > Date: Thu, 29 Apr 2010 10:24:28 +1000 > From: Nathan Kroenert <[email protected]> > To: Tony MacDoodle <[email protected]> > Cc: [email protected] > Subject: Re: [ldoms-discuss] Performance Bad on Virtual Disks > Message-ID: <[email protected]> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Can you characterize the performance of the system? > > Does the application do a lot of SYNC writes? > Lots of small reads? > Lots of seeks? > > There are a bunch of tools available for looking at interesting ZFS > sorts of things. > > I'd highly recommend Richard Ellings zilstat script. It'll help to show > you if you are backing up on a lot of SYNC writes to the ZIL. > > IIRC, (and I don't trust my memory these days) just about everything > that's written down to the IO domain is treated as SYNC and if all you > have is a couple of disks backing it, that might throttle you to about > 300 write IOPS (or 600 at best with a stripe) and if your IO's are tiny, > then your bandwidth will be tiny. > > I'd suggest that it would be a good idea to run some of the DTrace > toolkit scripts such as iosnoop and iotop to get a feel for the size and > number of IO's your webserver is giving you. > > I can tell you that a well configured T-series is an ideal WEB platform > - though I more frequently use containers than LDOMS when it comes to > workloads that are short-sharp in and out style apps. (Any v11n layer > can add to latency for individual transactions... Containers are the > thinnest. :) > > Let us know what you turn up - and for the fun of the exercise, (and to > allow you to rule in or out the ZFS component), try giving the LDOM a > raw slice and see how things go. > > Cheers! > > Nathan. > > > > Tony MacDoodle wrote: > > T5140's, LDom 1.3, Solaris 10 u8, ZFS for both backend (RAID 10, > > internal 146G 10K rpm) and rpool...... Horrible performance for a web > > server.... We are thinking on not using the product............ > > > > On Wed, Apr 28, 2010 at 6:24 PM, John <[email protected] > > <mailto:[email protected]>> wrote: > > > > We're seeing similar performance problems as well. What are the > > details of your configuration, ie what kind of disk's, what storage > > array, etc > > -- > > This message posted from opensolaris.org <http://opensolaris.org> > > _______________________________________________ > > ldoms-discuss mailing list > > [email protected] <mailto:[email protected]> > > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > > > > > > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > ldoms-discuss mailing list > > [email protected] > > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > > > > ------------------------------ > > Message: 8 > Date: Wed, 28 Apr 2010 19:00:30 PDT > From: John <[email protected]> > To: [email protected] > Subject: Re: [ldoms-discuss] Performance Bad on Virtual Disks > Message-ID: <1453433040.481272506460151.javamail.tweb...@sf-app1> > Content-Type: text/plain; charset=UTF-8 > > Are you seeing the bad performance from the IO/Control Domain as well, or > only from inside Guest Domains? > -- > This message posted from opensolaris.org > > > ------------------------------ > > _______________________________________________ > ldoms-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > > > End of ldoms-discuss Digest, Vol 33, Issue 13 > ********************************************* > -- Thanks & Regards, Arup Das 9903966741
_______________________________________________ ldoms-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
