Tim Jones
Forum Replies Created
-
Hi Bob,
The issue isn’t the Areca-powered mTape unit. The Areca mTape firmware and OS X driver (it’s the combination) for that specific chipset has been tested and in fact, part of the Thunderbolt approval involved both Apple and Intel testing the unit with BRU PE. That unit works properly as we worked with mLogic and Areca to get it to where it is.
The frustrating result is that even though the Areca team now have the answer to the tape compatibility issue, they’re not expanding the changes to include their general population RAID HBAs.
They’re also apparently aware of the heat profile issues as it was their USA rep that recommended the fan solution to us 3 years ago.
I’m hoping that we hear more from Suny and his results with the new BRU PE build and using alternative source drives for the tests to fort her isolate the Areca array from the equation.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
Hi David,
In this case, the Areca HBA is a separate controller. In our lab environment, the Areca fell over when it was ONLY used to control the DASD.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
[Neil Sadwelkar] “BTW how did this data get into the Areca in the first place. If it copied this huge amount of data successfully, then the unit seems to be working fine.”
Actually, while that seems to make sense, it’s not a good measure of stability. When copying data to and from the filesystem, applications read and write data in chunks and the filesystem manager is caching even larger segments of data for burst writes and reads. Also, the low level drivers utilize memory management capabilities known as “scatter gather segmentation” to provide both improvements to observable performance and memory utilization. If you tracked actual drive I/O, you would see that this allows the drive to be quiesced for large periods relative to the actual application use. A quick test copying 300GB of data from one array to another at over 400MB/sec showed disk activity in the 42% average range (use cycle). This is the same mechanism that LTFS-based solutions use since they are running through the FUSE filesystem manager.
When BRU is reading data from the drives, we are using a much more direct API and there is no caching or scatter gather operations. Using BRU and reading that same 300GB of data from the array resulted in a disk use cycle of over 88%. This is also creating a much higher level of I/O stress on the RAID controller chipset. Combining these two heat generation situations will easily overdrive the Areca chipset’s heat profile limits. This is why simply adding a fan blowing over the cards resolved the crashes in our lab.
There is a very big difference in heat generation between 42% and 88% usage cycle for 8 spinning hard drives.
For the record, we have never seen these types of crashes / hangs with ATTO or HighPoint RAID HBAs or chipsets.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
Kiki,
That command is correct, just keep in mind that there a lot of things that can go wrong using a simple “cp” command with tape as the destination. This is because “cp” has no way to deal with tape-related issues that may occur.
Just like you weren’t “born” to your editing tools and had to learn them to be proficient, I recommend that you spare a bit of time to learn more about the Unix command line tools available on your Mac if you’re going to enter this realm. I recommend this book to everyone looking to get a bit more familiar with the power that the command line brings to your Mac:
A Practical Guide to UNIX for Mac OS X Users by Mark Sobell. It provides a very easy to follow background to the Terminal.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
Suny,
It could very well be their enclosure. Since you’re running with 8 drives, there’s a lot of heat generated – especially during a backup / archival operation where you’re not simply copying files. In our tests, we were able to burst write or burst read from an Areca-controlled array, but when we started continuous and contiguous reads from the controlled array, the HBA would overheat and the Mac wold either hang or fully gray-screen.
As contrived as this may sound, if you have a small desk fan, you might try placing it in front of the case (open the front door if there is one) and blow air across the drives to reduce the heat that they output. This may be enough to reduce the heat profile within the case and keep the Areca controller under critical temps. It’s what Areca’s tech rep suggested to us when we reported the issue and it did work when the fan was on medium or high speeds.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
One additional bit of information that our lab team just shared with me – the Areca RAID cards are known to have heat problems when backups are running. Unlike normal disk I/O when you are writing data in bursts, the act of reading data from the device for backup is a continuous stream of data with very few breaks (especially with LTO-6). He have had our Mac Pro and Windows systems physically shut down because the Areca card overheats. When we shared this issue with Areca US support, their guidance was to open the case and aim a desk fan at the system.
While it is unfortunate and something that I don’t like to have to do, we now recommend to our customers to stay away from Areca RAID adapters for non-server installation as they are simply not stable in extreme I/O situations where you are not running them in a controlled server infrastructure environment (forced cooling, humidity control, etc.).
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
Hi Suny,
Unfortunately, all that we can offer is to insure that you’ve downloaded their latest firmware and drivers. Unlike ATTO and HighPoint, the Areca team has been very strict with their work in resolving the tape I/O issues that we identified in their driver / firmware stack over 5 years ago. It wasn’t until they discovered that their drivers didn’t work for the mTape unit that they even tried to make any effort in resolving the issue.
I have contacted the mLogic team. In the mean time, our engineers are building a version of BRU PE that may provide a more detailed examination as to why the data flow is so slow. Check your support ticket for a download link later today.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
That is definitely not right. Please make sure that you are running 3.1.10 (152) and open a support ticket at support.bru.com.
As for the verify, we complete the backup portion and then execute the verify pass on the completed archive. This way, you can abort the verify and re-run it later without affecting the entire backup operation.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
What do you get if you click the “How Many Tapes” button above the “Start Archive” button?
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters! -
Hi Enrique,
Using the generic LTFS freebies will get you to where they are asking you to be. Just be aware that you need to use the LTFS version for your vendor’s device (you can’t use HP’s LTFS for IBM). Also, if you opt for an LTO-6 drive instead of the LTO-5, you can still create the LTO-5 tapes that they are asking for.
Tim
—
Tim Jones
CTO – TOLIS Group, Inc.
https://www.tolisgroup.com
BRU … because it’s the RESTORE that matters!