Watching Oracle Disk Manager work with Solaris dtrace
Hi, I had to get ODM to work in a Solaris 10 container (aka zone) and wanted to make sure that Oracle did indeed use ODM and direct I/O to the Oracle files. I could not find documentation for the Veritas dtrace probes but after some trying and guessing the following dtrace script turned out to be helpful. #!/usr/sbin/dtrace -s #pragma D option quiet fbt:odm:odm_vx_*io:entry /!self->trace/ { self->trace = 1; } fbt:vxfs:vx_dio_rdwri:entry, fbt:vxfs:vx_dio_check:entry /self->trace/ { printf("%s %s\n", probefunc, stringof(((struct vnode *) ((struct vx_inode*) arg0)->i_vnode)->v_path)); } fbt:odm:odm_vx_*io:return /self->trace/ { self->trace = 0; } The vx_dio_check probe seems to trigger whenever IO takes places via ODM. When the IO is not cached, i.e.via Cached Oracle Disk Manager, the vx_dio_rdwri probe also triggers. Example: vx_dio_check /zones/ora/root/orcldata/bodm1/control02.ctl vx_dio_rdwri /zones/ora/root/orcldata/bodm1/control02.ctl vx_dio_check /zones/ora/root/orcldata/bodm1/control01.ctl vx_dio_rdwri /zones/ora/root/orcldata/bodm1/control01.ctl vx_dio_check /zones/ora/root/orcldata/bodm1/system01.dbf vx_dio_rdwri /zones/ora/root/orcldata/bodm1/system01.dbf vx_dio_check /zones/ora/root/orcldata/bodm1/example01.dbf vx_dio_check /zones/ora/root/orcldata/bodm1/example01.dbf vx_dio_check /zones/ora/root/orcldata/bodm1/control01.ctl vx_dio_rdwri /zones/ora/root/orcldata/bodm1/control01.ctl I had caching enabled for example01.dbf via CODM so we don't see the vx_dio_rdwri probe for the exampl01.dbf, only the vx_dio_check probe. This is for Storage Foundation 5.1SP1 Solaris x64. Perhaps someone finds this useful. Generally my impression is that ODM is a very useful performance feature that most people have licensed but which nevertheless is seldom used.
Apache monitor in Linux RHEL 6.2 with SF 5.1SP1
Upgrading Linux RHEL 6.1 - using sfha-rhel6_x86_64-5.1SP1PR2RP2 on a VCS cluster, has a problem with the Apache monitor script, I checked in the bundled agents guide for a debug mode. The Apache mount point is a shared resource and started by VCS after the CreateMntPt has verified the mount point does not exist. The default value for the CreateMntPt is 0 but it seems that under VCS , the file system should not exist, therfore it should be set to 2. The engine_A.log is reporting on both nodes about the filesystem does not exist, how do I change the monitor script to stop checking for the file system on both nodes at the same time?
