Fujitsu D2616 issues (and LSI SAS9201-8i question)

[foolishlywise]

I am attempting to kill two birds with one stone on this thread -

I ended up with a D2616 (based on the SAS 2108) which has 512mb cache after a server screwup. Stupidly I thought the cache would be a good idea so I put all 7 drives onto it and shoved it into the server. Used the (painful) MegaRAID BIOS thing to get the thing to boot and yeah. The current setup is having everything as RAID0 volumes. Not too bad since SMART data passes through -

However, I have no idea what is going on but this setup decides to sometimes write at 100MB/s, then 50MB/s and then drop to 0B/s - for example, writing to a WD Green 2TB starts off 100MB/s and then decides to just drop 40MB/s (on a totally solid archive file). Sometimes, it just tells me that an 'unspecified error is keeping me from copying this file'.

Everything is set to off - cache is disabled on ALL logical volumes because I don't have a UPS yet and whilst power outages are rare (once a year or something), I can't risk it.

All in all, is this a limitation of using a RAID card like the D2616 (which has to be used IR with Fujitsu firmware and everything RAID0)? Would I be better off switching to a SAS9201-8i? (The 8i is the most basic non-cache pre-IT flashed card)

Cheers

 

[pricklypunter]

If you are using software based raid, then an HBA or a RAID card that can be made to look like one, is what you need. No fancy cache or other abstraction etc, let your software based disk controller handle it all. The closer you can get your disks to your raw hardware the better. In this case, you won't go wrong with any of the LSI2008 based cards floating around, IBM M1015, DellH200/ 220/ 310 etc. If you need 12Gbps capability, for SSD for example, any of the 3008 based cards like the LSI9300-8i etc will do you nicely

Oh and issues like you are describing are usually disk controller timing issues of one kind or another, could be a failing disk, or just one that's slow in responding to your RAID card or even a heat or power issue. It's never really a good idea to marry enterprise capable controllers with consumer level disks. Even if it works most of the time, invariably it will bite your a$$ at some point

 

[foolishlywise]

Awesome, thank you!

I've got two H310's on the way and intend to keep the 9201-8i solely for the tape drive in the system.

Would anyone know if the timings on the H310 (IT mode on LSI's fw) can be adjusted through the config utility?

 

[pricklypunter]

If by timings you hope to be able adjust out any issues with your disks, I'm afraid you're out of luck. The timing issue is down to the individual disk not responding in time for some reason. It's all down to the disk mechanics, associated disk controller and the disk firmware, the problem is not on the HBA/ RAID card

Software based RAID implementations are usually a hell of a lot more forgiving of poor timing than hardware based solutions, so you might find that once you swap over your cards and allow your software RAID to take full control, that your issues melt away. Assuming of course you don't actually have anything mechanically going on with your disks

 

[foolishlywise]

I think it could be a symptom of using WD Green drives in a storage pool (eco-friendly, spin down a lot) so when a file is written to a disk, the software thinks there is an issue (whilst the drive is spinning up) hence is like nope cant copy. That could explain why the issue happens on BOTH controllers (the Fujitsu as well as the LSI).

Cheers for your help, I'll post my findings as soon as I can