Forum Replies Created

Viewing 1 - 10 of 77 posts
  • Alex Gardiner

    February 16, 2019 at 10:25 am

    Could it be that the QNAP reports space in TiB?

    Just a thought for the weekend.

    72TB is pretty much 65TiB 🙂

  • Alex Gardiner

    January 28, 2019 at 4:31 am

    To echo bobs post, use double parity.

    Single parity with 10TB disks is not a good idea. Once you end up in a degraded state with HDDs this big, the chances of failure during a rebuild are uncomfortably high.

  • Alex Gardiner

    January 19, 2019 at 1:41 am

    [Razvan Dimitriu] “Can the initialization process be the cause for the slow speeds or maybe I’m missing something else?

    If the ATTO is anything like your LSI, user generated IO will push initialisation into the background. Even then, benchmarks are going to be impacted until init completes. On an LSI you can usually set the “BGI Rate”.

    Either way I would suggest running the test again once the process completes…

    Remember that without init you can’t be sure your drives are in a good state. Same goes for Consistency Checking (ATTO call this ‘media scan’). I would suggest scheduling this fortnightly for early on Saturday morning. To an extent the frequency depends on type of drives used/capacity/working hours.

  • Alex Gardiner

    December 24, 2018 at 9:40 am

    [Bob Zelin] “Don’t do big data transfers while you are trying to edit. This applies to any system, from any manufacturer. “

    Many people underestimate this fact. +1

  • Alex Gardiner

    November 30, 2018 at 9:22 am

    EDIT: I woke up and realised I was thinking about the Myricom driver – this *does* set the interface to jumbo frames by default. Mellanox probably won’t.

    Either way check you don’t have a mismatch, it’ll cause headaches.

  • Alex Gardiner

    November 29, 2018 at 10:03 pm

    Yep, check MTU – the mellanox driver defaults to MTU 9000 in some environments, seen that before.

    Ifconfig or ethtool on the linux side if poss, see what it’s saying 🙂

  • Alex Gardiner

    October 16, 2018 at 7:50 am

    It can be a little tricky to update the firmware on these. You’d need a bootable USB drive running Free DOS, or something similar. The flashing utility itself is easy to use, but perhaps best left for a rainy day when you have nothing else to do.

  • Alex Gardiner

    October 15, 2018 at 7:23 pm

    I have a hunch the 9240 is going to do better than on-board RAID.

    What I’m not really sure about is how much processing the SAS2008 really does. My impression is that it probably hands work over to the host CPU, which on that generation of card is probably not a good thing.

    Then again, if you are really going for RAID0 (which is simple, but very unsafe), then you might get results that are good enough for your needs.

    Overall I’d say Bob is right, you are best off looking for a newer generation of card.

    The other thing to do is to just try what you have. The 9260 will probably give you better results and access to RAID6, but it’s also an EOL card as well.

  • Alex Gardiner

    October 14, 2018 at 9:44 pm

    From memory the 9240 used the SAS2008 chipset while its bigger brother (the 9260) used the SAS2108.

    The 2108 was a far more powerful RAID on Chip (ROC) and had 512MB of cache. The 2008 could do some RAID levels, but it was better suited to HBA type duties and was cache-less. FWIW you’ll see the SAS2008 cropping up in cards like the 9211 (and IBM based clones), which could be flashed with a special firmware that simply passes drives over to the host (IT mode). This is popular with ZFS users because the card doesn’t interfere with commands sent to the disk, which is exactly what you want.

    Modern equivalents are any HBA running the 3008 chipset (for HBA duties) or the 3108 for MegaRAID.

    So is the 9240 a real hardware RAID?

    Kind of and not really, all at the same time…. and definitely not as much as the 9260 ☺

    …for legacy hardware it really depends what kinds of speeds you want to hit. The 3Ware 9750 was based on the same chipset as the 9260 and was the mainstay of shared storage products for years. LSI eventually acquired 3Ware’s assets, which was later acquired by Avago and finally assumed into Broadcom.

    One thing is for sure, as Bob says, the Areca cards are popular and have firmware that is especially suited to video streaming (3Ware had a similar thing going on for years, but I’ve never been especially clear if those ideas were assumed into the MegaRAID stack). Atto had their own ideas about this too.

    In the end I expect hardware RAID will be phased out in favour of smarter more flexible equivalents – filesystems such as ZFS can kind of do it all, but that is a discussion for another day.

  • Alex Gardiner

    September 9, 2018 at 1:41 pm

    Tentatively chipping in…

    Scrubbing will push the system hard because it has to read every sector on every HDD in order to verify integrity. So like Bob says that could be highlighting a weak PSU.

    You could also have a HDD issue if the machine chucks the towel in after about the same uptime. It depends how bad it is, but I’ve seen storage go nuts if a drive is seriously acting out of turn. Doesn’t matter if it’s ZFS, hardware RAID, MD – this can cause horrible problems. Some systems are just better at this than others at this.

    Could also be thermal or memory, but both seem unlikely – especially if your system uses ECC.

    If you can get in to the QNAP via SSH try running the dmesg command, in particular try with the timestamp switch eg. dmesg -T. You might be able to see the kernel buffer throwing messages about a particular drive that has yet to fully fail.

    NB: I have no idea if dmesg is installed on a QNAP, but it is a Linux system, so I expect so.

    Some more info here…

    As for scrubbing, I’d schedule this at least monthly as a weekend task. Thats my view, feel free to ignore.

Viewing 1 - 10 of 77 posts

We use anonymous cookies to give you the best experience we can.
Our Privacy policy | GDPR Policy