Hi Shai, This is a shot in the dark, but can you confirm that the Sun networking is working correctly? I have had problems in the past with Solaris connecting to the network in half-duplex mode. Here are a couple of commands that might help:
1. Use dmesg to scan for boot-time messages about networking, e.g.: 'dmesg | grep duplex' 2. Use ndd to check the link mode e.g., 'ndd -get /dev/hme link_mode' I noticed from recent posts on the net that the half/full duplex problem seems to exit even in Solaris 10 as well as of course earlier releases. We're working on commercial support for Solaris with uni/cluster. If I think of more things as we go through certification I'll be sure to post them to this list. Thanks, Robert On 5/22/08 10:48 AM, "Shai Weinstein" <[EMAIL PROTECTED]> wrote: still no luck with the Solaris.. any ideas ? Thanks On Fri, May 16, 2008 at 10:19 AM, Shai Weinstein <[EMAIL PROTECTED]> wrote: Hi I even tried to load the recommended patches for solaris nad the problem still accurs its not only the long time that it takes to bring the backend up, but it totally hangs (and not reacting to new client connections at all -not refusing connection, just not responding) for JVM u mean this ? bash-3.00# /usr/jdk/j2sdk1.4.2_13/bin/java -version java version "1.4.2_13" Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_13-b06) Java HotSpot(TM) Client VM (build 1.4.2_13-b06, mixed mode) Thanks On Thu, May 15, 2008 at 12:00 PM, Emmanuel Cecchet <[EMAIL PROTECTED]> wrote: Hi Shai, Thanks a lot for this feedback. There is certainly a problem with the JVM for Solaris. My guess is that this probably comes from the network stack implementation and how TCP_NODELAY can be implemented on that platform (hence the longer recovery time). Could you try another JVM on Solaris to see if you notice any improvement? What is the exact vendor and version number of your Solaris JVM? Thanks for the feedback, Emmanuel Ok So I went and tried the same setup with a difference: original setup: 1 controller on solaris 10 (sparc), with 2 linux mysqld backends. now: 1 controller on linux with 2 linux mysqld backends I tested and I didn't have the problem of the controller hanging.. and controller keeps getting connections while backing up/enabling a backend + enabling the backend doesnt take too long like with a solaris controller... it seems to hang right after a backup is finished and it try to re-enable backed up backend what could be the solaris problem ? (I use same java 1.4.2 for the controller and 1.5.0 for myositis on both OSs) Thanks On Tue, May 13, 2008 at 12:42 PM, Shai Weinstein <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]><mailto:[EMAIL PROTECTED]>> wrote: hello On Tue, May 13, 2008 at 1:12 AM, Emmanuel Cecchet <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]><mailto:[EMAIL PROTECTED]>> wrote: Hi Shai, it was not disabled before the backup.. also no errors was displayed during or after backup Did you see a message saying that the backup was completed? What was the backup state when you tried to enable it? this from the log when i start the backup: http://pastesite.com/794 the backend was automatically enabled but during this time (~3 minutes) the vdb was not accesible by any client (its not in production so only 1 client..) on Myosotis i set the persistent connections option to false I don't know what else to try , maybe the mysql connector is not ok ? i'm using the latest stable one from mysql.com <http://mysql.com><http://mysql.com> <http://mysql.com><http://mysql.com> version 5.1.6 there are no any errors, only this warning: 12:26:46,970 WARN DatabaseBackend.myDB.one Default connection manager undefined in backend configuration, setting to a VariablePoolConnectionManager I dont think I use persistent connections - here is my config (in a paste site): http://paste.lisp.org/display/60634 The config looks good. The persistent connection option is set on the driver side so I can't tell from the virtual database config file. I have not played with persistent connections for a long time, so there might be a bug that induces the behavior you indicate but this looks strange since we can log open/close persistent connection events. So it should not block anything even while enabling a new backend. Try to see if it changes anything when persistent connection are enabled or disabled. You might need help from the Myosotis mailing list if you don't find the option there. Note though that your connection pool size is only 50 per backend so you can't have more than 50 client connections. After 50 connections new clients will be blocking no matter what is the state of your backends. Which version of Sequoia are you using? 2.10.10 This is good. Keep us posted with your progress, Emmanuel -- Emmanuel Cecchet FTO @ Frog Thinker Open Source Development & Consulting -- Web: http://www.frogthinker.org email: [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]><mailto:[EMAIL PROTECTED]> Skype: emmanuel_cecchet -- Robert Hodges, CTO, Continuent, Inc. Email: [EMAIL PROTECTED] Mobile: +1-510-501-3728 Skype: hodgesrm
_______________________________________________ Sequoia mailing list [email protected] https://forge.continuent.org/mailman/listinfo/sequoia
