This is kinda quirky. Somtimes my guests would hang after ntp started. With some playing around I determined this happens only in odd numbered domains. Even numbered domains run fine but when ntpd starts in the odd domains existing ssh connetions hang, and I can''t make new ssh connections. The afflicted domain does respond to ping and to ''xc_dom_control.py shutdown''. This happens if I run the guests serially using the same root partition. Domain 0 is not running ntp. I suppose the clock is a shared resource so I ought to only have DOM0 set it, and not the guests. This is with the xeno-1.2.bk tree. As for HIGHMEM4G, I am interested in working on it, but it will be a couple weeks before I have an opportunity. ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > This is kinda quirky. > > Somtimes my guests would hang after ntp started. With some playing around > I determined this happens only in odd numbered domains. Even numbered > domains run fine but when ntpd starts in the odd domains existing ssh > connetions hang, and I can''t make new ssh connections. The afflicted > domain does respond to ping and to ''xc_dom_control.py shutdown''. > > This happens if I run the guests serially using the same root partition. > > Domain 0 is not running ntp. I suppose the clock is a shared resource > so I ought to only have DOM0 set it, and not the guests. > > This is with the xeno-1.2.bk tree.Is the box a dual? It could be an SMP-related issue -- odd domains would run on CPU1 by default. Running NTP on non-privileged domains should work -- by default a non-privileged domain will sync against Xen (and Xen is in turn synced against ntpd in DOM0). If a non-priv domain runs an NTP daemon then it essentially detaches itself from Xen wall-time. -- Keir ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
" Is the box a dual? It could be an SMP-related issue -- odd domains " would run on CPU1 by default. Aha, yes it is a dual. And, sure enough, I can start/stop ntp fine on CPU0, then pin the domain to cpu1, and hang it by starting ntp. ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > " Is the box a dual? It could be an SMP-related issue -- odd domains > " would run on CPU1 by default. > > Aha, yes it is a dual. And, sure enough, I can start/stop ntp fine on CPU0, > then pin the domain to cpu1, and hang it by starting ntp.Is this on Redhat 9? I can have a go at reproducing it tomorrow -- if I can do that then it shouldn''t be that hard to fix. -- Keir ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
" Is this on Redhat 9? I''m running debian/testing. That particular guest was not entirely up-to-date, running ntp-4.0.99g / libc-2.3.1 So I updated it to ntp-4.1.2a / libc-2.3.2 and the problem has gone away. Now I can restart ntpd on either cpu. ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> I''m running debian/testing. That particular guest was not entirely > up-to-date, running ntp-4.0.99g / libc-2.3.1 > > So I updated it to ntp-4.1.2a / libc-2.3.2 and the problem has gone away. > Now I can restart ntpd on either cpu.Bizarre. We mostly use RH9 which is ntp-4.1.2 and libc-2.3.2. It''s hard to understand why the CPU the domain is running on makes a difference with ntp-4.0.99. There''s obviously a Xen/Xenolinux bug that this particular versions tickles... If you''ve got a serial line connected, please can you try hitting ''q'' when you''ve got a couple of domains in the dead state. That should at least tell us whether they''re looping or waiting for an event. It might be worth doing the following to see if the system calls are any different between the two ntp versions, and also to see what the system call before the hang is: strace ntpd [args] >/dev/console 2>&1 Cheers, Ian ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel