Here is a, hopefully, growing collection of scripts to help manage VxVM disk storage subsystems. All scripts are GPL, and written in PERL.
Thanks to a very very generious individual, 2 JNI HBA's have been donated. Furthermore, 2 SCSI diffs were as well, which means that when I scrape together some cash for an AIT/DLT drive, I can do NetBackup tutorals! I am eternally grateful.
This is a small PERL script which gathers data together about your disk subsystem. It is intended (and useful only for) systems relying on A5x00 (aka: Photon) disk arrays. The problem with using the arrays is this: you commonly use these disk arrays for high speed, redundant disk systems, based on RAID0+1 or RAID1+0 layouts. In these cases you commonly will create 2 (or more) fibre channel arbitrated loops and mirror volumes disk for disk. In order to manage this type of storage design you must name each vmdisk something equivelent to it's location, therefore you carefully name your vmdisks something like "arrayA_r0", to designate that this disk is in disk array "A", and it's in Rear Slot 0. This is a great way to manage your disks. But, what happens when a failure occurs? There are two sparing methods: relocd and sparecheck. Sparecheck will spare whole disks, whereas relocd will spare out subdisks one-by-one, very messy. The real problem being that when you use sparecheck the vmdisk that spares out the failed disk will be renamed to whatever the failed disk was named, therefore when you look at your "vxprint" output without looking at the SCSI IDs you won't notice that anything is wrong... however in reality you may now have a volume in which both mirrors are using disks in the same disk array, breaking the redundancy that we built in.
So, what I really needed was a tool that could tell me all of the following things:
The point being that I can run the simple script, get a nice list to print out and then make sure that the VMDisk name matches the disk location. Other info, like the driver instance number is nice for the records. The previous method of doing this was to sit down with about 100 pages of output from: path_to_inst, vxdisk list, vxprint, luxadm display X, and maybe even the listing of /dev/dsk, and then spend hours putting together a disk map. But this takes ages, and is prone to error. I think it's a good exercise for new admins, and a useful exercise for older admins, but when you come back from a 2 week vacation and can't determine exactly what has changed and need to manaully verify the config, you don't wanna wait to finish a map.
It should be noted that other tools, such as jtplex and STORtools fall short of the mark, but are close alternatives. I had little to no luck with jtplex, ontop of the fact that it didn't list A5x00 disk locations. STORtools, on the other hand, makes wonderful disk maps, but is unaware of the VM requiring you to still do a manual disk map, but with most of the hard work done for you. Besides this, STORtools should NOT be used lightly or frequently, and certainly not in critical production enviroments when there is never a good time. Therefore I built this script, which tries to be as non-intrusive as possible. It is run per photon so you can spread out the runs... we all know what happens when you run luxadm too much or too closely to a previous run, SES gets confused and cranky. If dumps the luxadm display X information to a /tmp file, the contents of the /dev/dsk/ directory to a /tmp file (much faster than polling the directory itself), and dump the vxdisk list output to a /tmp file, and then runs regex's on these tmp files. When it's done it removes these files.
Future updates: I'm thinking of adding at least one more feature to the script, other than cleaning up my ugly code. I want it to run thru the vxprint information to tell me which volumes the disk is a part of and what the subdisks name is. I'll be writting that in soon.
Your all familar with the standard disclaimers, and the restrictions of the GPL. Redistribute but credit cuddletech and/or myself. Change it all you like, but if you make some nifty changes let me know. Etc, etc. If you loose your entire subsystem cause of my script it's your own damned fault, don't blame me. And lastly, read it before you run it. I don't trust other peoples scripts, I don't expect you to trust mine. Do note, that you must run the scripts as root, since only root has perms to run luxadm.
|lux_parse.pl v1.0||Oct 10th, 2002: Initial release, bad looking code, debuging stuff left in, bad variable names, etc.|
|lux_parse.pl v2.0||Oct 11th, 2002: Cleaned up code, column format, more expandable, handles most types of errors from photon.|
Here is an example of v2.0's output. Several errors have been simulated (*yoink!*) to test it:
bash-2.03# ./lux_parse.pl.v2.0 cuddlestor Cleaning up: ... done. Drive Listing for A5100 named cuddlestor: A5x Slot Short WWN Status SCSI ID Instance VMDisk DG -------- ---------------- ------ --- cuddlestor,f0 ERROR Off(Bypassed:AB) - - - - cuddlestor,f1 20372d0f69 On (O.K.) c1t1d0s2 ssd12 cuddle-f2 cuddledg cuddlestor,f2 20370971e8 On (O.K.) c1t2d0s2 ssd10 - - cuddlestor,f3 2037097752 On (O.K.) c1t3d0s2 ssd3 cuddle-f3 cuddledg cuddlestor,f4 20370970f3 On (O.K.) c1t4d0s2 ssd0 cuddle-f4 cuddledg cuddlestor,f5 20370d44c8 On (O.K.) c1t5d0s2 ssd7 cuddle-f5 cuddledg cuddlestor,f6 ERROR Not Installed - - - - cuddlestor,r0 20370e0b08 On (O.K.) c1t16d0s2 ssd9 cuddle-r0 cuddledg cuddlestor,r1 20370e85f8 On (O.K.) c1t17d0s2 ssd13 cuddle-r1 cuddledg cuddlestor,r2 20370d3bef On (O.K.) c1t18d0s2 ssd5 cuddle-r2 cuddledg cuddlestor,r3 20370d44ee On (O.K.) c1t19d0s2 ssd8 cuddle-r3 cuddledg cuddlestor,r4 203714322b On (O.K.) c1t20d0s2 ssd2 cuddle-f0 cuddledg cuddlestor,r5 20370971df On(No UNIX Label) c1t21d0s2 ssd6 - - cuddlestor,r6 20370d44d2 On(No UNIX Label) c1t22d0s2 ssd11 - - Done...Cleaning up: vxdisk.tmp.out, luxadm.cuddlestor.out, devdsk.tmp.out... done. bash-2.03#
You'll notice that in the above example I've gotten v2.0 to adiquately deal with failure conditions. I'll test against other conditions when I find a new way to break my array, but these are the 3 most common: Disk ByPassed, Disk Not Installed, UNIX Label fault (ie: vtoc is crap).
Next is a more practicle example, of a system with 6 photons (I'll only show 1), all A5200's using DMP, etc,etc. As you can see, this photons layout could be cleaned up a bit, and thats the point of it all. Again, this is v2.0 output:
bash-2.02# ./lux_parse.pl.v2.0 a Cleaning up: ... done. Drive Listing for A5200 named a: A5x Slot Short WWN Status SCSI ID Instance VMDisk DG -------- ---------------- ------ --- a,f0 20374fe451 On (O.K.) c2t0d0s2 ssd424 arraya01 prod1gr c4t0d0s2 ssd58 a,f1 2037935ec6 On (O.K.) c2t1d0s2 ssd563 arraya02 prod1gr c4t1d0s2 ssd564 a,f2 20379742b3 On (O.K.) c2t2d0s2 ssd565 arraya03 prod1gr c4t2d0s2 ssd566 a,f3 2037199aca On (O.K.) c2t3d0s2 ssd413 arraya04 prod1gr c4t3d0s2 ssd45 a,f4 203719dda1 On (O.K.) c2t4d0s2 ssd405 arraya05 prod1gr c4t4d0s2 ssd32 a,f5 20374fe161 On (O.K.) c2t5d0s2 ssd393 arraya06 prod1gr c4t5d0s2 ssd16 a,f6 203719d9a4 On (O.K.) c2t6d0s2 ssd402 arraya07 prod1gr c4t6d0s2 ssd29 a,f7 203719d392 On (O.K.) c2t7d0s2 ssd415 arraya08 prod1gr c4t7d0s2 ssd47 a,f8 203719dce9 On (O.K.) c2t8d0s2 ssd414 arraya09 prod1gr c4t8d0s2 ssd46 a,f9 203719dc8e On (O.K.) c2t9d0s2 ssd421 arraya10 prod1gr c4t9d0s2 ssd55 a,f10 203719de70 On (O.K.) c2t10d0s2 ssd429 arraya11 prod1gr c4t10d0s2 ssd0 a,r0 203719d7d2 On (O.K.) c2t16d0s2 ssd420 arraya12 prod1gr c4t16d0s2 ssd54 a,r1 203719d868 On (O.K.) c2t17d0s2 ssd408 arraya13 prod1gr c4t17d0s2 ssd38 a,r2 203719dcc6 On (O.K.) c2t18d0s2 ssd412 arraya14 prod1gr c4t18d0s2 ssd42 a,r3 2037935ec1 On (O.K.) c2t19d0s2 ssd572 arraya15 prod1gr c4t19d0s2 ssd571 a,r4 203719d02f On (O.K.) c2t20d0s2 ssd406 arraya16 prod1gr c4t20d0s2 ssd36 a,r5 203719d905 On (O.K.) c2t21d0s2 ssd404 arraya17 prod1gr c4t21d0s2 ssd33 a,r6 2037936d80 On (O.K.) c2t22d0s2 ssd580 arraya18 prod1gr c4t22d0s2 ssd579 a,r7 203719dcab On (O.K.) c2t23d0s2 ssd417 arraya19 prod1gr c4t23d0s2 ssd50 a,r8 203719d76b On (O.K.) c2t24d0s2 ssd411 arraya20 prod1gr c4t24d0s2 ssd41 a,r9 203719d961 On (O.K.) c2t25d0s2 ssd396 arraya21 prod1gr c4t25d0s2 ssd23 a,r10 203793836f On (O.K.) c2t26d0s2 ssd586 arraya22 prod1gr c4t26d0s2 ssd585 Done...Cleaning up: vxdisk.tmp.out, luxadm.a.out, devdsk.tmp.out... done. bash-2.02#
|For questions or comments mail email@example.com|