what did you learn today?

Jump to latest Follow Reply
Status
Not open for further replies.
Muted (0)
Whittey

Whittey

Ars Tribunus Militum
2,006
  • #6
    That Jimmy Johns is freaky fast only when ordering for a family. An office-sized order is not freaky fast. It was still freaky good.

    McAfee Agent 4 is significantly more reliable than 3.6, but still isn't even close to "good enough."

    Setting permissions on cifs shares on an openfiler box with a good sized AD is a 30minute process. Which sucks.


    -=Whittey=-
     
    S

    SideAngleSide

    Ars Scholae Palatinae
    1,472
    Subscriptor
  • #8
    Originally posted by chalex:
    2) Getting "kernel: Machine check events logged" in your /var/log/messages on your x86_64 machine? Use 'mcelog' to get more info:
    http://www.mjmwired.net/kernel.../x86_64/machinecheck
    Click to expand...

    mcelog + kdump = Everything I need to know when hardware fails.

    Today, I learned that NFSv4 is much easier to support than NFSv3 (so long portmapper). I can now very easily tunnel NFS connections over SSH for systems that don't support SSHFS.
     
    Barmaglot

    Barmaglot

    Ars Legatus Legionis
    11,783
    Subscriptor
  • #9
    Originally posted by ghub005:
    As an example - Adaptec currently ship 12 SAS RAID controllers of which only 3 have firmware that supports more than 8 drives (these are hard limits, drive multiplexers aren't supported by those cards either).
    Click to expand...

    Their website says otherwise. Pretty much every controller they currently ship (which is way more than 12) supports either 128 (for series 2 and 3) or 256 (for series 5) devices. The only ones that don't support expanders are RAID 4000 and series 1 controllers (the latter are based on SiI chips, I think).
     
    G

    ghub005

    Ars Tribunus Angusticlavius
    8,702
  • #10
    Originally posted by Barmaglot:
    Originally posted by ghub005:
    As an example - Adaptec currently ship 12 SAS RAID controllers of which only 3 have firmware that supports more than 8 drives (these are hard limits, drive multiplexers aren't supported by those cards either).
    Click to expand...

    Their website says otherwise. Pretty much every controller they currently ship (which is way more than 12) supports either 128 (for series 2 and 3) or 256 (for series 5) devices. The only ones that don't support expanders are RAID 4000 and series 1 controllers (the latter are based on SiI chips, I think).
    Click to expand...

    Brain fart! I meant to write LSI not Adaptec.
     
    B

    bombcar

    Ars Legatus Legionis
    32,602
  • #14
    Originally posted by Barmaglot:
    Originally posted by ghub005:
    Brain fart! I meant to write LSI not Adaptec.
    Click to expand...

    Ah, LSI is another story, I'm not familiar with them at all as they aren't sold in Israel except as Dell and Intel rebrands, and Dell PERC5/PERC6 do support expanders.
    Click to expand...

    We had an Adaptec SAS card that would support 24 drives but could only build 12 drive RAID packs.

    I also learned that next day HP support is next day, is in, it arrived before I did.
     
    S

    sjg

    Ars Scholae Palatinae
    958
  • #30
    That VMware Converter is awesomely awesome.

    Used it a couple of times before on healthy, non-critical stuff. Client had a very sick win2k/ex5.5 server with all sorts of config oddities (was effectively a one-node cluster, for one thing). It had corrupted it's information store and had real problems with it's logfile and DB arrays - just copying a few small files about would take forever, sometimes hanging it either temporarily or permanently. IS far too big to be fixed with eseutil or isinteg, and restoring it from tape would fail after a couple of hours at horribly slow speeds.

    They'd had some disk failures in the past but the controller was quite happy with the state of things, and no relevant events to say something was amiss. No spare controllers to try either.

    Fortunately they had an ESXi box which we'd prepared for another project. Put on Converter, set the ESXi box as the target, a little over an hour later it's virtualized self was running perfectly. Restore worked great, IS came up fine, the world is saved.
     
    D

    duffbeer703

    Ars Scholae Palatinae
    636
  • #33
    Originally posted by montegard:
    I learned that sending people who just don't "get it" to training usually just means you've spent some money and they still don't "get it".
    Click to expand...

    I've observed five strategies that are used to isolate the inept from the productive where firing isn't an option.


    • Develop tedious, repetitive tasks for them to perform. They'll either do them and leave you alone or find alternate employment.
    • Create an "Island of Misfit Toys". Concentrate the inept together, away from productive people.
    • Give them nominal responsibilities with no impact. "Customer Liason", "Junior Project Manager", "Senior Statistician", etc
    • Put them in charge of service requests.
     
    S

    scorp508

    Ars Legatus Legionis
    10,214
  • #34
    Originally posted by duffbeer703:
    I've observed five strategies that are used to isolate the inept from the productive where firing isn't an option.


    • Develop tedious, repetitive tasks for them to perform. They'll either do them and leave you alone or find alternate employment.
    • Create an "Island of Misfit Toys". Concentrate the inept together, away from productive people.
    • Give them nominal responsibilities with no impact. "Customer Liason", "Junior Project Manager", "Senior Statistician", etc
    • Put them in charge of service requests.
    Click to expand...


    1. Then the union steps in and says you're penalizing them.

    2. We have one and its fun to watch, but unfortunately they run a system that the rest of us depend on and it ends up affecting us.

    3. Ugh.... position bloat.

    4. Then they can't handle the SRs and the productive ops people step in to do the work anyways.

    I really wish firing was easier. -- :( --
     
    Z

    Zaphod

    Ars Legatus Legionis
    16,903
    Subscriptor
  • #35
    Originally posted by scorp508:
    Originally posted by duffbeer703:
    I've observed five strategies that are used to isolate the inept from the productive where firing isn't an option.


    • Develop tedious, repetitive tasks for them to perform. They'll either do them and leave you alone or find alternate employment.
    • Create an "Island of Misfit Toys". Concentrate the inept together, away from productive people.
    • Give them nominal responsibilities with no impact. "Customer Liason", "Junior Project Manager", "Senior Statistician", etc
    • Put them in charge of service requests.
    Click to expand...


    1. Then the union steps in and says you're penalizing them.

    2. We have one and its fun to watch, but unfortunately they run a system that the rest of us depend on and it ends up affecting us.

    3. Ugh.... position bloat.

    4. Then they can't handle the SRs and the productive ops people step in to do the work anyways.

    I really wish firing was easier. -- :( --
    Click to expand...



    I take it that the fifth way is to have them assassinated by ninjas.
     
    T

    tinyMan

    Ars Tribunus Angusticlavius
    6,149
    Subscriptor++
  • #38
    ZFS is as cool as it is made out to be.

    We have a host that has 4x 2gb FC connections to a disk array. When reading or scrubbing that array we see damn near line-speed:

    Code: 
    zpool iostat 10                       
                 capacity     operations    bandwidth
pool           used  avail   read  write   read  write
------------  -----  -----  -----  -----  -----  -----
disk_storage  19.6T  39.2T  6.09K      0   779M      0
disk_storage  19.6T  39.2T  6.12K      0   783M      0
disk_storage  19.6T  39.2T  6.15K      0   786M      0
disk_storage  19.6T  39.2T  6.11K      0   782M      0

    Just reading from the array (a dd to /dev/null) nets 780 MB/s, just 20 MB/s under the speed we achieved when directly accessing the underlying block devices directly. And very, very close to the theoretical max speed you can get using 4x 2gbs fiber paths.

    Code: 
    zpool status
  pool: disk_storage
 state: ONLINE
 scrub: scrub in progress, 0.25% done, 8h2m to go


zpool iostat 10
                 capacity     operations    bandwidth
pool           used  avail   read  write   read  write
------------  -----  -----  -----  -----  -----  -----
disk_storage  19.6T  39.2T  5.47K      0   691M      0
disk_storage  19.6T  39.2T  5.31K      0   671M      0
disk_storage  19.6T  39.2T  5.46K      0   690M      0

    The scrub, which is actually doing some processing on the data is proceeding at ~690 MB/s, which is way above what I expected out of the anemic CPUs in the T2000 we are using for the testing.

    The best part (so far) is that it whipped up a ~60TB filesystem on a bunch of devices in 5 seconds.

    Code: 
    zpool list
NAME                    SIZE    USED   AVAIL    CAP  HEALTH     ALTROOT
disk_storage           58.8T   19.6T   39.2T    33%  ONLINE     -
     
    Z

    Zhengyi

    Ars Praetorian
    454
    Subscriptor
  • #39
    I learned that when you move an entire Helpdesk & IT Ops group downstairs, away from every other person in the company, two things happen:

    1) There's a hell of a lot fewer walkups and interruptions, so suddenly there's a lot more work getting done, and

    2) There's a hell of a lot more people filing tickets instead of just walking up and asking for something, so suddenly upper management is a bit more receptive to pleas for additional hands.

    Oh yeah, and sometimes a complete reset of your desk environment is just the ticket to finally getting it properly (re)organized.
     
    Status
    Not open for further replies.
    Jump to latest Follow Reply
    Jump to latest Follow Reply