Network performance degrades after a few days

  • This is my first OMV build and it's running fairly well except that the network performance degrades to about a 1/10 bandwidth after a few days. I'm **not** losing the network, and it's **not** always slow, just after some period of time. On a fresh boot, these are my benchmarks:


    * iperf = 116 MB/s
    * pv/smb = 104 MB/s
    * rsync = 86 MB/s


    I'm happy with that! However, when this problem occurs, they all drop -- so I think it is protocol independent.

  • I don't believe there's anything amiss, the server is underloaded -- the most I do is copy files over and server a single (non-transcode) Plex stream in the evening. I've noticed this when copying files after it's been running for a few days. Next time it happens I'll give a more complete reckoning from glances. Presently, glances reports the cores are at 47C, CPU at 25%, MEM at 11%, SWAP at 1.2%, 15m load at 0.43, SYS DISK is 4/20G.

  • I ran my quick diagnostic (dd if=/dev/zero bs=1G count=1 | ssh omv 'cat > /dev/null') and noted it is taking for ever, so while it was running I took the screenshot above. Everything in glances looks okay.


    When my diagnostic completes it reports:


    1073741824 bytes (1.1 GB, 1.0 GiB) copied, 913.378 s, 1.2 MB/s


    iperf reports the equivalent in bits (9.41 Mbits/s)

  • Why do you say that? The system has 8GB of RAM, which should be plenty and is barely used.


    ```
    ╭─reagle@omv ~
    ╰─➤ vmstat -s
    7985784 K total memory
    733520 K used memory
    2811316 K active memory
    1325020 K inactive memory
    3289004 K free memory
    517048 K buffer memory
    3446212 K swap cache
    8206332 K total swap
    19968 K used swap
    8186364 K free swap
    370049 non-nice user cpu ticks
    3789 nice user cpu ticks
    174475 system cpu ticks
    103285356 idle cpu ticks
    49982 IO-wait cpu ticks
    0 IRQ cpu ticks
    34187 softirq cpu ticks
    0 stolen cpu ticks
    8184847 pages paged in
    10099879 pages paged out
    280 pages swapped in
    2582 pages swapped out
    46973006 interrupts
    73248057 CPU context switches
    1575744089 boot time
    94451 forks
    ```

    • Offizieller Beitrag

    So it sounds like there is something wrong with the suspend procedure. Some driver that is not correctly initialized after suspend, perhaps. Or even defective hardware.


    I have no idea how to find out what might cause it, except replacing hardware one piece at a time and reinstall from scratch until it works as it is supposed.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!