Find the answer to your Linux question:
Results 1 to 3 of 3
Enjoy an ad free experience by logging in. Not a member yet? Register.
  1. #1
    Just Joined!
    Join Date
    Jul 2014
    Location
    Bay Area, California
    Posts
    96

    High number of packet discards


    Hello,

    So we recently did a datacenter consolidation on a highly compressed schedule (we weren't planning on it)

    Since the consolidation we've been having some issues. We've believe we've finally found a possible cause in a high number of disconnects. We're running streaming video & users get disconnected often etc & servers lose connection to internet.

    Here's a small glimpse into some of what we're seeing

    discards.png

    We're seeing traffic peak at about 10Mbs. Solarwinds (which we just setup as a trial, probably going to purchase) is marking lo interfaces as 10Mbs(loopback interface).I ran ethtool across the eth0 & eth1 (since lo interface doesn't return configs) and they all report at 1000Mbs.

    I'm uncertain what could cause it to report as 10Mbs. How would I look into this?

    I'm unsure how to view the loopback interface speed or if there is a configuration somewhere that might tune this... Possibly a kernel parameter since lo is a kernel module.

    It's reporting 100% lo use at 10Mbps.

    Any help would be much appreciated!!

  2. #2
    I doubt your issues are related to the lo interface - from Google for "solarwinds loopback 10mb"

    So, this is a loopback? You can't really set it to a static speed, the OS dynamically adjusts the speed to accommodate the applications; which is why I (and most network folks) typically don't monitor them for stats. However, if its a SQL server loopback, I do monitor them but don't alert on them; and only look at them if someone is having issues with SQL latency. You'll probably notice NPM has discovered the the interface as a 10MB speed as well (because it couldn't determine a real speed), and depending on how heavy the server is used, this is usually enough (unless there is an issue).
    Your info seems to be from the switch side, so you need to isolate if this is a switch or NIC issue - or the mix of switch and NIC model (they don't always play nice together.)

  3. #3
    Just Joined!
    Join Date
    Jul 2014
    Location
    Bay Area, California
    Posts
    96
    -->
    Quote Originally Posted by HROAdmin26 View Post
    I doubt your issues are related to the lo interface - from Google for "solarwinds loopback 10mb"



    Your info seems to be from the switch side, so you need to isolate if this is a switch or NIC issue - or the mix of switch and NIC model (they don't always play nice together.)
    Hey thank you so much for the quick response. I actually was just about to post that I figured out it's not the loopback for sure and I think it's from VLAN misconfigurations causing it to discard packets from improperly tagged VLANs it's not assigned to. I believe it's something our network team is going to have to fix...

    However, just to share with everyone how I proved it's not the loopback interface. On a system already taking live traffic (so that means the speed would be higher if it was idle) I performed the following:

    [user@hostnameredacted (puppet) (08:48:1 ~]# nc -d -l 9999 > /dev/null & time dd if=/dev/zero bs=1M count=10000 | nc localhost 9999
    [1] 1853
    10000+0 records in
    10000+0 records out
    10485760000 bytes (10 GB) copied, 34.3119 s, 306 MB/s
    [1]+ Done nc -d -l 9999 > /dev/null

    real 0m34.314s
    user 0m2.517s
    sys 0m58.406s

    So this shows I'm writing at 306MB/s which is well over the 2Gbps I need with my two 1Gbps ports which never run maxed out anyways...

    So quick explanation for anyone who can't decipher it.

    I used netcat to open a listening port that would not be taking input from STDIN & dumping any incoming traffic to /dev/null.

    I then launch a timed dd outputing from /dev/zero with a size of 1MB sending a total of 10000. It then computes the passage of data & the time it took to copy that 10GB across the loopback interface.

    I'll post back when a solution is found whether it's a linux solution or not.

$spacer_open
$spacer_close

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •