OT: Cache settings on RAID's? (Bob...)
Hey, Bob. I posted this question a while back, but never heard a response. You mentioned in your review something about changing the cache settings on all RAID's from their default. I'm just about to set up my 3rd Sonnet Fusion, this time with 5-750 Gig drives. Anything I need to know about the cache setting? I usually just do Raid-0 using Apple's disk utility software...usually works fine. (I have had one drive go down before...not too bad for heavy use over the course of a year).
I almost never do uncompressed HD, so changing the block size is not that important to me, from the default setting of 32K. But if you want maximum performance, set it to 128K (you click on OPTIONS before you create the RAID in the RAID tab). With that said, the current Sonnet X4P or E4P will not perform as fast as the Cal Digit FASTA-4x or FASTA-4e.
You have a great setup, and if you do nothing, it will continue to work just fine.
Bob, we're going to be doing a project with 4 SD Uncompressed 8 bit streams multicam. Would changing the block size make any difference for this scenario? What is the downside to changing it to 128k? There must be a downside or wouldn't they just ship them like that?
FWIW, I just re-striped a RAID on my second FCP system (still on Mac OS 10.3.9) and there was no option in the disk utility for block size. I checked on my main FCP (which is 10.4.8) and there was an option, so if you have an older setup you might not get the choice of block size.
First excuse me, but my questions/comments are sort of related to this post.
I have a G5 dual 2.5/ 8gb memory, FCP 5.1.2, and Kona 2 card, osx 10.4.8. I built a raid a few weeks back consisting of the Fusion 500p enclosure, Sonnet Tempo SATA X4P, and 5 x Maxtor 7H500F0 Sata II 500 gb drives.
Here is the weird problem: all works well when I stripe 4 of these drives with mac disk utility. I stripe at 0 raid, 256k option, and all works fine. It is when I try and add the 5th drive that I run into all sorts of problems.
The raid seems to stripe correctly, I copy all my files (from a backup firewire drive) sometimes my files copy correctly, other times they freeze at various points. When I do finally have all my footage copied over I restart FCP and run the timeline. As I run the time line it freezes and sometimes get the beachball endlessly spinning. At other times my raid dissapears for no reason.
I have tried restriping, no go. I have returned the 5th drive and gotten it replaced, no go. It is definatley not the drives, as I have striped the raid a variety of ways with all sorts of combinations, in example with the 3 original drives and new replacement drive...and it works just fine. In other words when I use 4 drives all is well, but with 5 it sucks.
I had originally checked for bad media. But why would bad mediawork on 4 drives, and not 5? Could I have a bad enclosure? so had it replaced, same problems. Maybe a bad bad controller card? The controller card was updated to the latest beta software, that did not help. Replaced controller card with a Caldigit and same problems.
I have tried a variety of diferent slots on both enclosures... no go.
Also tried ALL the various cache and striping settings as well (including using SoftRaid).
Called Sonnet tech dept, however not too helpful, suggestion wise, short of returning all hardware, which I did. Caldigit had me try some tests reseting Pram, testing memory, clearing login items, fcp prefs etc. No Luck.
Anyone ever run into these issues, or similiar??
Any info is greatly appreciated.
Please excuse me for accusing you of things that you may have tried already.
First, let me assure you that the Sonnet and Cal Digit 5 drive chassis work just fine. If you replaced the "bad drive" and the problem was still there, it is my opinion that you had a bad chassis. You said that you replaced the Sonnet Fusion 500P, and had the same problem, but I find this hard to believe. There is ABSOLUTELY no need to replace the host controller card, the cable or the software to stripe that you are using (Apple Disk Utility works just fine).
The bottom line here is
1) you have a bad drive
2) you have a bad enclosure (drive box)
3) you have no idea of what you are doing
First, you have to be confident that you can erase all the media on these drives, as we are going to do lots of testing.
1) put all 5 drives in the system and boot up.
2) Go into Apple Disk Utility
3) delete any RAID group that you may have created (do this in the RAID tab)
4) ONE AT A TIME, erase (initialize) each of the 5 drives - as a MAC OS-X
Extended volume - NEVER choose Journaled, or anything else. When you get done, you should have FIVE seperate drives on your desktop.
5) Do you have 5 drives mounted on your desktop - because if you don't, you either have a bad drive, or a bad slot
6) now, click on Disk First Aid in the Apple Disk Utility, and run this on EACH of the 5 drives - did it work ? Did each drive pass
7) run AJA Kona System Test on EACH of the 5 drives - you can use the minimum test of 786x486 or whatever the default test is, at 128M file size.
Test EACH DRIVE. You should get over 50mb/sec with EACH DRIVE minimum.
Does this work ? If you don't get over 50mb/sec from EACH DRIVE (one at a time), the slow drive is your problem drive
8) if this does work, go back into APPLE DISK UTILITY, and click on the RAID tab, and drag all the drives into the center box, and create a STRIPPED RAID as MAC OS X Extended - DO NOT select Journaled.
Did it work ?
9) now, click on Disk first aid, and test the SATA RAID - did it work
10) now, go back to AJA Kona System test, and do the same test on this RAID, but pick a higher rate, like 1920x1080 and a 1 Gig file size. You should
see results of over 180mb/sec, even with a 32K block size, which is the default for the Apple Disk Utility.
FOLLOW EVERY STEP THAT I POSTED HERE. and report back your results. If you don't do this, I will not help you.
Hey, Bob. Every time I think I know what I'm doing, you throw a little morsel of wisdom out there.
Why not "Journaled" under RAID Settings? I'm pretty sure that's the way I've done my past 2 RAIDS and they work fine, but am I taking a performance hit or something? I have this brand new 5x750gig RAID that I just set up. Nothings on it yet, so let me know if I should re-initialize it with "Extended" but not "Journaled."
Thanks a ton. Good stuff here...
here is the honest answer - I forgot why "not journaled". I was in some basic Apple XServe RAID training with Apple, and they made a big issue to NEVER use Journaling for the Xserve RAID, or you will get poor performance. This applies to other video media arrays as well. Your performance will decrease. For the true technical answer (which I was told explicitly by Apple) - I forgot ! I don't remember ! but I know they (Apple) said NEVER do it - just for your boot drive in the G5.
If your server contains high-bandwidth usage data files, such as large video, graphics, or audio files, you may want to weigh the benefits of using journaling against the performance needed to access your data. In most cases, the impact of journaling upon data access performance are unnoticeable to users, but its implementation may not be practical for servers where data access demands outweigh its benefits.
If a volume contains read-only data that is not mission-critical, it may not be necessary to turn on journaling if performance is more important than safety.
Journaling is best suited for servers requiring high availability, servers containing volumes with many files, and servers containing data that is backed up at infrequent intervals (nightly, for example).
When journaling is turned on on a storage volume, the server automatically tracks file system operations and maintains a continuous record of these transactions in a separate file, called a journal. The operating system can use the journal to return the file system to a known, consistent state after a failure. This eliminates the need to perform a consistency check on the entire file system during startup. Instead, when the server is restarted, Mac OS X simply replays recent transactions in the journal, bringing the system up-to-date and resuming operations that were interrupted during the failure. With a journaled file system, server restart takes just a few seconds, regardless of the number of files, or the size of the volume.
Thanks to everyone for your help on my raid adventures. I have been building FCP, Avid, and Media 100 systems for over a dozen years, so I am not complete NEWBEE when it comes to raids, formatting, etc. To be honest this is my first 'port multiplier" raid so I am a bit stumped as to why it has not gone smoothly as hoped. I did try all the steps mentioned by BOB weeks ago, with no luck. However I did go through all of Bob's steps again "to the last detail" and sad to say, no luck. I DO HAVE (2) of the Fusion enclosures at the moment (the original suspect, and the replacement) and both are giving me the same results.
All drives were erased one by one, tested in Disk Utility, and Kona Test, they all passed and I obtained between 62 to 72mb/sec read results. I tried to stripe and it failed. Started over again, erased disks, ran tests (all ok), striped, and this time it worked. Received read rates at close to 230mb/sec.
Now I copied all files (288gigs) from fw800 drive to new 2.3gb striped volume and it seemed to copy ok. When I tried to access files on this new volume, as I played a QT file, I recieved a device removal message. Basically I am recieving the same errors that have been stumping me for weeks. For giggles I tried onced again a 4 drive raid and behold it works.
I will be calling Sonnet on Monday to see if there may be any other items I need to review.
As I mentioned earlier I had suspected the original enclosure to be faulty, and now my replacement enclosure seems to have a rather tight slot (the lowest) where the drive (any of them) does not seem to slide in as easily as the others. Possibly I may have lucked out with TWO bad enclosures....who knows?
This is my "guess", and it is only a guess. I will probably regret guessing this, but I believe that you may have TWO bad Fusion chassis. Nothing else makes sense.
The following statement is also my opinion. I don't know how much Sonnet will be able to diagnose for you over the phone. You have done every test, and you have a "second" chassis. Sonnet builds a box that holds Hitachi or Seagate drives, and they buy the port multiplier chip in the box from Silicon Image. They are not the "ground zero" designers of this stuff. Lets face it - if you sent someone a replacement box, and they said that one was bad too, you probably would not believe them. What is happening at your place makes no sense, because you can initialize EACH DRIVE, and get the correct read and write speeds on EACH DRIVE, but stripping them makes it fail with 5 drives (not 4). It's got to be 2 bad boxes. It is my suggestion that you send your boxes back to Sonnet. Until you do, they will probably think that you are crazy.