[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-users] Xen Migration Failure
Thanks. Your problem is definitely the difference of the hardware capabilities of the boxes. I ran into the same thing when I built my cluster. Short answer is that you must always start your domain on the box with the most common caps, which in your case is Server02. Long answer is below. Basically, when a domain starts it remembers the capabilities from the host on which it started. You'll even see it in the kernel boot messages. After that, anything with less capabilities will not run the domain. It will complete the migration, but as soon as the destination host tries to start it, it will immediately die. Take a look at the hw_caps line from both boxes: Server01: bfebfbff:00100000:00000000:00000180:0000441d Server02: 3febfbff:00000000:00000000:00000080These are bit masks, encoded as base64. If we take the caps strings and convert them to binary, we get: Server01: 101111111110101111111011111111110000000000010000... Server02: 001111111110101111111011111111110000000000000000...They're actually much longer than that, but that's enough to get the idea across. As you can see, Server01 has two bits set that Server02 doesn't have. A domain started on Server01, when migrated to Server02, will try to continue to use the capability it had when it was on Server01, and because Server02 can't do that, Server02 will immediately error and destroy the domain. However, notice that all the capabilities that Server02 has, Server01 also has. So if you start a domain on Server02, when migrated to Server01, it will not error, because Server02 can do everything Server01 could. And because the capabilities are remembered at boot time, the domain will not notice that the host it's running on now has more caps than it did before. So you can migrate it back to Server02 and it will work fine. This is how I believe it works, but I could certainly be wrong. If there's someone with more knowledge about what the hw_caps are and the way they affect PV domains, please share what you know. -Jeff On Jun 23, 2008, at 7:55 AM, Cody Jarrett wrote: Sure, here is the info: Server01: root@localhost:~ $ xm info host : localhost release : 2.6.18-53.1.21.el5xen version : #1 SMP Tue May 20 10:31:46 EDT 2008 machine : i686 nr_cpus : 2 nr_nodes : 1 sockets_per_node : 1 cores_per_socket : 1 threads_per_core : 2 cpu_mhz : 3000 hw_caps : bfebfbff:00100000:00000000:00000180:0000441d total_memory : 502 free_memory : 2 xen_major : 3 xen_minor : 1 xen_extra : .0-53.1.21.el5 xen_caps : xen-3.0-x86_32p xen_pagesize : 4096 platform_params : virt_start=0xf5800000 xen_changeset : unavailable cc_compiler : gcc version 4.1.2 20070626 (Red Hat 4.1.2-14) cc_compile_by : mockbuild cc_compile_domain : cc_compile_date : Tue May 20 09:27:25 EDT 2008 xend_config_format : 2 Server02: $ xm info host : server02 release : 2.6.18-53.1.21.el5xen version : #1 SMP Tue May 20 10:31:46 EDT 2008 machine : i686 nr_cpus : 1 nr_nodes : 1 sockets_per_node : 1 cores_per_socket : 1 threads_per_core : 1 cpu_mhz : 1694 hw_caps : 3febfbff:00000000:00000000:00000080 total_memory : 1023 free_memory : 128 xen_major : 3 xen_minor : 1 xen_extra : .0-53.1.21.el5 xen_caps : xen-3.0-x86_32p xen_pagesize : 4096 platform_params : virt_start=0xf5800000 xen_changeset : unavailable cc_compiler : gcc version 4.1.2 20070626 (Red Hat 4.1.2-14) cc_compile_by : mockbuild cc_compile_domain : cc_compile_date : Tue May 20 09:27:25 EDT 2008 xend_config_format : 2 Attachment:
smime.p7s _______________________________________________ Xen-users mailing list Xen-users@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-users
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |