Oracle Exalogic vServer Hung – Running xm dump-core

Have you been tasked with capturing a dump of a VM? Are you running an Oracle VM implementation? Are you using an Oracle Exalogic?

Here I’ll show you what options we have to use and my recommended methods to provide the data to support (assuming that is why you are attempting the capture of the dump-core).

First we’ll need to capture the GUID

[root@mwlab01-ovs01 crash]# xm list
Name ID Mem VCPUs State Time(s)
0003gt01232900009cee514c12330a14 46 6144 6 -b---- 36291.5
0003gt0123290000a1b9042a9c6d45a9 60 65536 16 r----- 2534.9
0003gt0123290000c6a86211f7df2181 28 32768 8 r----- 579200.1
0003gt0123290000cc0222c2fcc2416c 37 12288 2 r----- 181463.7
Domain-0 0 5632 24 r----- 2621106.4

Now that we have the guids of our running vms, lets ensure that it matches our machine name. For this example we will use the GUID of 0003gt0123290000a1b9042a9c6d45a9

[root@mwlab01-ovs01 crash]# grep -Ri 'simple' /OVS/Repositories/*/VirtualMachines/*/vm.cfg |grep -i '0003gt0123290000a1b9042a9c6d45a9' |awk '{print $3}'

Onto the dump-core. Here are the available options

[root@mwlab01-ovs01 crash]# xm dump-core
Usage: xm dump-core [-L|--live] [-C|--crash] [-R|--reset] [Filename]Dump core for a specific domain.
-L, --live Dump core without pausing the domain
-C, --crash Crash domain after dumping core
-R, --reset Reset domain after dumping core

Now that we know the options, lets create the dump-core

[root@mwlab01-ovs01 crash]# xm dump-core -R 0003gt0123290000a1b9042a9c6d45a9 /ssd/var/crash/mwlab01-vm01-vmcore
Dumping core of domain: 0003gt0123290000a1b9042a9c6d45a9 ...

You’ll see the file size is huge. Its a sparse file so it will shrink considerably using tar for easier upload to support

[root@mwlab01-ovs01 crash]# ls -lh
total 65G
-rw------- 1 root root 65G Dec 2 07:54 mwlab01-vm01-vmcore
[root@mwlab01-ovs01 crash]# tar -zcf mwlab01-vm01-vmcore.tar.gz mwlab01-vm01-vmcore
[root@mwlab01-ovs01 crash]# ls -lh
total 66G
-rw------- 1 root root 65G Dec 2 07:54 mwlab01-vm01-vmcore
-rw-r--r-- 1 root root 1.2G Dec 2 08:09 mwlab01-vm01-vmcore.tar.gz

Fairly simple process and something any level of sysadmin can perform. The file is now very small and can easily be provided to support. This could even be automated as a script from 1 admin server to go out and find what node the machine is running on.

Leave a Reply

Your email address will not be published.