Results 1 to 6 of 6

Thread: Slow Attachment upload, slow sending messages with attachments

Hybrid View

  1. #1
    Join Date
    Aug 2013
    Rep Power

    Default Slow Attachment upload, slow sending messages with attachments

    Folks, let me preface this as my first post and apologize if I'm in the wrong area of the forums.

    Since installing Zimbra, our FOSS version 7 install was uploading attachments very quickly. The dedicated physical server has a public IP and internal IP, 16gb ram, 10k mdadm raid 5 disk, a 6 core desktop processor. Bandwidth available is 5 down / 0.768 up. There are about 45 active users at any time.

    Message attachment is taking a very long time and sending a message with an attachment is also taking a long time.

    Things tired (and the result)
    - firefox and chrome using public IP and private IP (no difference)
    - html and ajax interfaces (slower to upload attachment on html, slow to send on either)
    - different workstations (no difference)
    - workstations without antivirus (no difference)
    - restarting zimbra (no difference)
    - restarting the server (no difference)
    - updated zimbra to the latest version (no difference)
    - tail -f /opt/zimbra/log/mailbox.log (don't see any issues during the attachment or sending process)
    - top (no significant memory or cpu spikes while attaching or sending)
    - jnettop (no significant bandwidth spikes during attach or sending)
    - free (memory is about 90% utilized)
    - df -h (disks have lots of space
    - disabling dns checks in the web UI Global / MTA settings
    - disabling rbl in the web UI Global / MTA settings
    - dmesg gave us some peculiar output (However, IOSTAT on the array indicates good disk throughput)

    [398765.293044] ata2: SError: { Handshk }
    [398765.293864] ata2.00: failed command: WRITE FPDMA QUEUED
    [398765.294680] ata2.00: cmd 61/a8:08:e2:f1:65/01:00:40:00:00/40 tag 1 ncq 217088 out
    [398765.294681] res 40/00:0c:e2:f1:65/00:00:40:00:00/40 Emask 0x10 (ATA bus error)

    Is there a zimbra admin out there who has experienced this?

    Thanks in advance.
    Last edited by boomit; 08-28-2013 at 11:31 AM.

  2. #2
    phoenix is offline Zimbra Consultant & Moderator
    Join Date
    Sep 2005
    Vannes, France
    Rep Power


    How about it being a HD controller problem


    Acompli: A new adventure for Co-Founder KevinH.

  3. #3
    Join Date
    Aug 2013
    Rep Power


    Thanks for the reply Bill.

    We are going to change the disk at ata2 and also all of the data ribbons tomorrow night. The board is a desktop model and I think the controller is: ATI Technologies Inc SB700/SB800 SATA Controller [AHCI mode] (rev 40)

    I'll follow up.

    Here are the results of iostat. sda, sdb, sdc are active in the raid 5. sdd is the spare disk to the raid 5. md0 is the partition with the message store on it.

    root@host:/home/user# iostat
    Linux 2.6.32-38-server ( 13-08-28 _x86_64_ (6 CPU)

    avg-cpu: %user %nice %system %iowait %steal %idle
    1.33 0.00 0.24 0.46 0.00 97.97

    Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
    sda 13.25 195.81 348.79 114714186 204343488
    sdb 13.32 187.91 358.87 110087688 210246960
    sdc 13.33 190.32 355.88 111497832 208497320
    sdd 0.00 0.00 0.00 1480 32
    md2 0.00 0.00 0.00 424 0
    md1 0.02 0.05 0.11 28904 65232
    md0 85.41 473.33 653.52 277305514 382866176
    sde 0.56 3.07 40.45 1801046 23698608
    Last edited by boomit; 08-28-2013 at 12:32 PM.

  4. #4
    Join Date
    Apr 2010
    Rep Power


    Sounds like a network issue to the server or maybe it's just the reality of only having 0.768 up. That doesn't sound like much but I'm used to Gb ethernet. :-)

  5. #5
    Join Date
    Aug 2013
    Rep Power


    I appreciate your thoughts and am looking into the MTA. Oddly it worked well until recently.

    The MTA was set to an external one. When I watch jnettop as the message is sending (slowly but surely) most of the upstream bandwidth is being consumed.

    I'm in the process of trying to change the MTA to localhost but am faced with the following error delivered in the form of a bounce-back message:

    Please turn on SMTP
    Authentication in your mail client.
    550-[]:33193 is not permitted to relay through this server
    without 550 authentication. (in reply to RCPT TO command)

    My thought is to use localhost to queue then deliver the message directly to the recipient without the need to rely on bandwidth provided by the external MTA.
    Last edited by boomit; 08-28-2013 at 02:51 PM.

  6. #6
    Join Date
    Aug 2013
    Rep Power


    It appears the issue is resolved.

    I changed the MTA to a localhost and turned on authentication. The web client now responds as expected in terms of upload speed. When hitting send, the web client sends the message quickly and I can see in jnettop that the message is being delivered directly to the mta for the receiving domain. I'll follow up with updates to SFP records.

    If I were to speculate a root-cause it would be that the relaying MTA had capped upload speeds, manifesting in a delay / hanging on the web UI while sending the messages.

    Thanks for the responses folks, much appreciated.
    Last edited by boomit; 08-28-2013 at 02:28 PM.

Similar Threads

  1. Replies: 0
    Last Post: 11-24-2010, 12:46 AM
  2. Slow to respond and attachments
    By mmedici in forum Administrators
    Replies: 3
    Last Post: 09-07-2010, 03:17 PM
  3. Slow Sending Messages with Zimbra?
    By Diranged in forum Administrators
    Replies: 5
    Last Post: 08-22-2010, 01:51 AM
  4. Download/View Attachment Very Slow
    By in forum Administrators
    Replies: 1
    Last Post: 05-29-2008, 07:24 AM
  5. Slow mails with attachments
    By qvindesland in forum Administrators
    Replies: 3
    Last Post: 11-16-2007, 05:24 AM

Tags for this Thread

Posting Permissions

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