Looking to update OmniOS/NAPP-IT from r151014

Discussion in 'Solaris, Nexenta, OpenIndiana, and napp-it' started by NOTORIOUS VR, Oct 23, 2018.

  1. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    I was getting some ldap console errors I believe was due to a time mismatch, so I changed that and rebooted. That said I saw an error again this time about boot time and rcache, and looking a little deeper I saw errors in the log about multipath status degraded on the disks?

    I've attached the log in txt format maybe that will help?

    startup-log.txt
     
    #21
  2. gea

    gea Well-Known Member

    Joined:
    Dec 31, 2010
    Messages:
    1,971
    Likes Received:
    639
    Mostly talkativeness

    Did you use SAS disks with mpio (You can disable mpio in Disks > Details > edit mpt_sas for LSI HBAs)?
    Did you use an AD that is not available during bootup? (does not matter for performance)
     
    #22
    Last edited: Jan 1, 2019
  3. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Hi Gea,

    my disks are 4TB WD RED's on 2x LSI cards.

    If the log isn't showing anything concerning to you then I am currently at a loss of what else to check of why I'm now having speed issues after the upgrade. I suppose I could try exporting and re-importing my pools into the old/original napp-it r15014 VM and see if something changes? What do you think of that?

    At least that would mean the napp-it instance can be ruled out and then the only thing left would be the ESXi upgrade.
     
    #23
  4. gea

    gea Well-Known Member

    Joined:
    Dec 31, 2010
    Messages:
    1,971
    Likes Received:
    639
    Problem finding can be hard as every installation is different.

    Especially from ESXi 5.5/ OmniOS 151014 to ESXi 6.7/OmniOS there are many changes, not only the OS but a switch to newer ESXi drivers and on OmniOS a switch from SMB 1 to 2.1

    As it seems iperf is ok as is the local ZFS performance as is the ESXi internal performance between VMs, the remaining options are between ESXi and its network stack to the Windows client.

    Have you tried another physical Windows client (one with a better nic than the Realtek)
     
    #24
  5. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Fair enough, if you believe the internal performance of the controller/disk/VM is correct then at least I can focus on what is left.

    Unfortunately... The only other client I can try is my laptop, but that would be using a TB3 adapter to GigE and I have no idea how good it is as I haven't needed to use it in this way (max throughput). My other option is to pull and use the 4 port Intel GigE NIC from my server that I am not using anymore since moving to 10GbE and install it into my desktop for testing.

    I will report back my findings when I have a chance to bring down the server to pull the NIC.

    thanks again for the help Gea, much appreciated.
     
    #25
  6. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Sorry for the delayed reply Gea. I tried my laptop and it showed the exact same limitations as my desktop.

    Again, my issue really cannot be my PC's or the network as that hasn't changed at all where I was always able to max out gigE speeds.

    That said, I've for now come to terms with it as I am currently building a new server to replace my existing one along with a new storage setup. Currently I have a 10x 4TB disk Z2 array and it has treated me well, but I am already past the recommended low disk threshold (just under 6TB free space).

    Currently I have about 16TB of Video Media, and this will be my primary focus for the new array. The remaining data is misc things that I care more about (ie. would like to have more safety) which is around 4.5TB right now and will grow less quickly. I don't currently run any VM's off of the array and if I would I would do so on SSD's not on the spindle drives.

    I'm currently trying to decide what I should do in regards to the new array. I have a 12 disk SA120 storage array for the new build (I should probably plan on getting a 2nd SA120 for expansion). I'm trying to wrap my head around this info: How I Learned to Stop Worrying and Love RAIDZ | Delphix

    Should I build another 10 disk array (with say 8 or 10TB drives) in Z2 and use the whole single pool for all my data agian, or should I split up to a smaller disk array in Z1 for media and make another Z2 array for my other files?

    I feel like I am limited either way I do it unless I add more disks. I don't want to worry about media space for some time.
     
    #26
  7. gea

    gea Well-Known Member

    Joined:
    Dec 31, 2010
    Messages:
    1,971
    Likes Received:
    639
    Without special reasons like power saving, performance or size too different, keep one pool on problems with the other etc, I would create one large pool from several z2 (6-10 disks) or z3 vdevs (>10 disks per vdev) as performance increase with disks and vdevs. With a lower overall fillrate, fragmentation is also lower.

    about LZ4
    Without compress, a ZFS block has a fixed size like 128k. Such a block (power of 2) has to be distributed over the disks of a vdev. Only when the number of datadisks is also a power of 2 this fits otherwise you have a waste what can lower overall capacity up to 10%.

    When you enable compress, size of a ZFS block is variable. As you mostly enable LZ4 today, you do not longer care about "golden numbers" of disks per vdev. With 24 disks I would use 3 x 8 disk Z2 vdevs
     
    #27
  8. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Sounds good, thanks for the suggestion (to calm my mind).

    That said, I think I will have to ditch my SA120 idea completely. It's just not condo friendly in the noise department :( Even my cat was swatting at near by switches (not plugged in but under the unit/table) because of the noise lol.

    After spending most of the day modding the Rosewill 4U to fit the new server EATX MB, I fired it up and flashed the on board 2208 to IT mode, connected the SA120 fired it up and was greeted with FAR too much fan noise for my use case from the twin PSU's fans (both PSU's were plugged in to get the fans to run at "normal" levels). I found a fan controller software for the SA120 but I think that is if it's being used with a JBOD controller not a pass-through controller (maybe - still have to play around a bit).

    So now I'm hunting for other cases I can use for my use case... but it seems like short depth (<18") cases are just not available. Norco's 24E is very expensive for just a JBOD chassis which is a shame.

    Worst case I will have to continue to use my existing tower case, but I really wanted to get things off the ground but with all the restrictions I have for space it makes my choices difficult (and expensive usually lol).

    [​IMG]

    [​IMG]
     
    #28
    Last edited: Jan 13, 2019
  9. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Gea,

    I'm attempting to get sg_ses to work under napp-it... not much info about it under omnios, I found one older Hard Forum thread where someone said it works under napp it but maybe I'm missing something... when I type sg_ses in the command line it says command not found?

    I'm desperately trying to control the fans on the SA120 to see if they can be at least bearable.

    Thanks!

    EDIT: I was able to regulate the fans down in a windows VM using the Lenovo utility as a test (gave up on napp-it for now for the fan test)... it's bearable but not sure if it's a real fix.

    Pulling the fan units out and the PSU I see they're using 40x40x30mm fans... Nocua has nice 40x40x20mm fans which would work nicely in the fan units but the PSU might be a little tricky since that would leave a gap between the chassis and the fan of ~10mm, that said some aluminium tape would be a quick fix to seal that area without too much trouble.

    Also some light reading here: https://forums.servethehome.com/index.php?threads/lenovo-thinkserver-sa120-rackmount-das.6829/page-6 tells me the wiring is non-standard. But that isn't a big road block for me.

    So 6x Noctua fans and some time I should be able to have a quiet SA120 in theory for around $125 CAD. Better than starting over I guess!

    [​IMG]
     
    #29
    Last edited: Jan 13, 2019
  10. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Hi Gea,

    I finally got around to exporting my pool from the latest instance of napp-it, and reimporting the pool into the previous instance (r151014) and guess what? The speed is BACK! I am again maxing out my gigE connections, and even transferring between the VMs/VMswitches is faster/more stable! No more speed issues! and I was able to rule out ESX 6.7U1 as well I suppose (although maybe there is an issue with it and your latest release?).

    Now the question is why? Would there be anything you require from me to help diag this?

    Transfer from file server to VM:
    [​IMG]

    Transfer from VM back to file server:
    [​IMG]

    Transfer from file server to Desktop (gigE):
    [​IMG]
    The little reductions in speed in the middle were from me loading this forum in a browser :p

    Transfer from Desktop to file server (gigE):
    [​IMG]
     
    #30
  11. gea

    gea Well-Known Member

    Joined:
    Dec 31, 2010
    Messages:
    1,971
    Likes Received:
    639
    I cannot confirm any problems with OmniOS 151028.

    So I would -
    - try another OmniOS (ex my ova template v4 if you use an older one, v4 comes with newest fixes)

    If this does not help
    - install OmniOS from iso and check with e1000
    - add vmware tools (pkg install open-vm-tools) and try vmxnet3s
     
    #31
  12. NOTORIOUS VR

    NOTORIOUS VR Member

    Joined:
    Nov 24, 2015
    Messages:
    43
    Likes Received:
    7
    Hi Gea,

    ok thank you.. in the coming weeks I will do that (probably on the new server I've built) and report back here. I'm quite certain that was using the current v4 OVA BTW.

    Thanks!
     
    #32
Similar Threads: Looking update
Forum Title Date
Solaris, Nexenta, OpenIndiana, and napp-it Looking for Atom C3x00+10Gb owner Nov 21, 2017
Solaris, Nexenta, OpenIndiana, and napp-it Looking to downgrade to netatalk 3.0.3 on napp-it / OmniOS Jun 25, 2013
Solaris, Nexenta, OpenIndiana, and napp-it Looking for Help - Napp-it Main Site Guide Jan 6, 2013
Solaris, Nexenta, OpenIndiana, and napp-it Omnios update error Mar 17, 2016
Solaris, Nexenta, OpenIndiana, and napp-it Help! unable to login after update attempt with Napp It Feb 15, 2016

Share This Page