I was reading posts about AutoRAID and not knowing much about them I wanted to make sure I was "treating" our unit correctly.
The unit is filled to capacity with 9GB SE drives, 3 fans/power supplies, dual 96MB controllers.
Since MirrorDisk/UX & OnlineJFS are not in the budget and we don't have a service contract on the machine I want to make sure I am doing everything I can to keep things running smoothly.
arraydsp output ---------------
Vendor ID = HP Product ID = C5447A Array serial number = 00000037F35D ------------------------------------- Array State = READY Server name = guppy Array type = 3 Mfg. Product Code = IJMTU00003
--- Disk space usage -------------------- Total physical = 104198 MB * Allocated to LUNs = 30624 MB * Used as Active Hot spare = 8683 MB * Used by non-included disks = 0 MB * Used for Redundancy = 17061 MB * Unallocated (avail for LUNs) = 47830 MB * -----------------------------------------
Currently I am not in any real need of more disk space. If I would need more space how much unallocated space should I leave for the RAID 0/1,5 automatic allocation?
Controller info ---------------
Information for controller X: Vendor ID = HP Product ID = C5447A Array serial number = 00000037F35D Primary = TRUE SCSI Address = 1 Serial number = R96LADH636281111 Paired serial number = R96LADH636701111 Product revision = HP56
Information for controller Y: Vendor ID = HP Product ID = C5447A Array serial number = 00000037F35D Primary = FALSE SCSI Address = 0 Serial number = R96LADH636701111 Paired serial number = R96LADH636281111 Product revision = HP56
My firmware is v.56. I think there are newer versions available. Is there any reason to update? I think the unit is setup correctly to use both X & Y controllers to maximize performance. I have seen Clay mention mount options for Oracle volumes that would increase performance with not additional concern about data loss.
My concern is that this box is not under any service contract. It is also a production machine so I am very afraid to make changes. :)
I have my make_tape_recovery tapes, Oracle hot backups. I am going to try and use dd to make a spare boot drive using another post I have seen from Clay.
The box is a K460/1GB RAM machine running Oracle 8.1.6 with a 4GB production database. We have been using the machine for just over a year and we will be growing about 5-6 times current size over the next 12 months.
You're a little behind on the firmware updates; the main danger is applying periodic patches. Many times the latest arraymgr software requires a firmware update or vice versa. However, since you are not under contract, I would be reluctant to update the firmware. Having said that, I have updated the firmware myself many times without incident. Man download for details. In my case, the boxes were covered so I didn't feel afraid to do it. I am rather amazed that you have a production box without maintenance. I have noticed that it is possible to buy a used 12-H with 2 controllers, 3 fans, and 3 power supllies (no drives) for about the cost of 1 year of 24x7 maintenance on the 12-H. That might be one option to give you a bit of security. For what it is worth, I've run AutoRAID's for many years and have never lost 1 bit of data.
It appears that you could add at least another 15GB's of LUN or so and notice no performance loss.
HP AutoRAID 12H uses "AutoRAID technology", which does automatic migration of data from RAID 0/1 to RAID 5 storage???Substantially reduces administrative overhead.
The HP AutoRAID array automatically retains the most frequently accessed data in high-performance RAID 1/0 and the less frequently accessed data in the most cost-effective RAID 5, tuning the storage subsystem real time. so you don't have to worry about how much unallocated space you should leave for RAID when you need more space.
For HP56 Firmware, It's O.K but recommend you update it to the latest firmware, HP62 to take adavantage of the latest features available like logging for data unavailabe events, improving performance during rebuiding.
You can afford to create another larger LUN. I'd go for 20GB, you really won't notice too much perf hit if data goes into R5.. In fact some data is best in R5.
I'd upgrade the controller f/w if I were you, just in ever.. I used to train on these guys and no matter how much students tried to screw up the f/w upgrade, by powering off, pulling out the controller, ctrl-cing, we could still not corrupt the autoraid.. just follow the procedure and you'll have no problem.
The autoraid is a reliable array, just be sure to shutdown correctly or you'll have major problems restarting, especially when you leave the autoraid powered off for more than 1 week.
ie: 1> umount /*12h_filesystems 2> vgchange -a n 2> arraymgr -s shut 3> shutdown -h now
This'll make sure all data is written to autoraid disk before shutdown/ power off and not stuck up somewhere in memory between the host and the disks.
I often pull out a disk from working autoraids just to see lights flashing when I'm in the lab.. users don't notice too much.. I blame windows when their mail goes slow hahahaa.. I do add the disk back again of course, and never pull more than one at a time! ;)
Autoraid, 12 disk array, 11 needed. 10 disk enclosure, 9 disks needed, 6 disk enclosure, 5 disks needed.. ie it's N+1.
Remember too, the loading order is top to botom, left to right. keep the disks near the fans.
