Page 1 of 2 12 LastLast
Results 1 to 10 of 13

Thread: Zimbra server errors, maybe configuration issue?

  1. #1
    jruiseco is offline Intermediate Member
    Join Date
    May 2011
    Location
    Colombia
    Posts
    22
    Rep Power
    4

    Default Zimbra server errors, maybe configuration issue?

    Hi,

    I've been using linux since redhat 4, and have been running servers with sendmail, dovecot and postfix for years now. The first thing is that I'm surprised is how slow zimbra is and the incredible amount of resources it needs. I still run a postfix server and it handles hundreds of users on a P4 1800 with 512MB of ram. That's why I think there must be something wrong with my config.

    The server:
    I'm running zimbra opensource with around 50 users on a box with 2 virtual Xeon X5650 cores and 4G of ram. The box only runs zimbra since it swallowed up all 4G of available RAM (around 38411728 usually). I benchmarked the box and it gave me well over 10,000 apache pages served per minute, so it's not the hardware (phoronix apache bench). IMAP threads are set to 500 from the web console, and pop threads are set to 300.

    My users complained so much about the webmail interface speed that they're now using outlook. They tried webmail, then zimbra desktop and now outlook (ugh!). The webmail speed even in text mode isn't the entire problem though. The console is slow as well. Is this normal? I've found zimbra tougher to configure than sendmail just because of the lack of information available.

    So what's in the logs? Here's the jist of today's errors. I'm kind of a newb to zimbra so I'm just looking at the mailbox.log. Can anybody explain what's going on?

    Thanks in advance,

    Joaquin

    [zimbra@crm ~]$ tail -35000f log/mailbox.log|grep "error\|failure"
    2011-05-02 06:28:06,367 INFO [Pop3Server-1641] [ip=186.99.232.92;] ProtocolHandler - I/O error while processing connection: java.net.SocketException: Connection reset
    2011-05-02 07:29:26,201 INFO [Pop3Server-1647] [ip=190.29.5.245;] ProtocolHandler - I/O error while processing connection: java.net.SocketException: Broken pipe
    com.zimbra.common.service.ServiceException: system failure: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    Caused by: com.zimbra.cs.index.queryparser.ParseException: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    com.zimbra.common.service.ServiceException: system failure: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    Caused by: com.zimbra.cs.index.queryparser.ParseException: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    com.zimbra.common.service.ServiceException: system failure: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    Caused by: com.zimbra.cs.index.queryparser.ParseException: Lexical error at line 1, column 83. Encountered: "\n" (10), after : ""
    2011-05-02 08:03:37,947 INFO [ImapServer-605] [ip=190.29.5.245;] ProtocolHandler - I/O error while processing connection: java.net.SocketException: Connection reset
    2011-05-02 08:16:50,833 INFO [ImapServer-611] [ip=190.29.5.245;] ProtocolHandler - I/O error while processing connection: java.net.SocketException: Connection reset
    2011-05-02 08:22:58,798 INFO [ImapServer-614] [name=nusuga@agsystems.co;mid=34;ip=190.29.5.245;] imap - S: 2krc BAD parse error: zero-length content
    2011-05-02 08:23:02,865 INFO [ImapServer-614] [name=nusuga@agsystems.co;mid=34;ip=190.29.5.245;] imap - S: x67q BAD parse error: zero-length content
    2011-05-02 08:23:03,379 INFO [ImapServer-614] [name=nusuga@agsystems.co;mid=34;ip=190.29.5.245;] imap - S: fcrx BAD parse error: zero-length cont

  2. #2
    Krishopper is offline Dedicated Member
    Join Date
    Dec 2006
    Location
    Minneapolis MN
    Posts
    777
    Rep Power
    9

    Default

    Java likes to consume memory and not return it, so the fact that it eats all 4GB of RAM shouldn't be an issue. I have 200+ users running on a box with 4GB of RAM, and it's quite fast.

    Is the machine digging into any swap?

    Also, you said "2 virtual Xeon X5650 cores" -- is this a virtual machine?
    01 Networks, LLC / Cybernetik.net
    Zimbra NE and OSS Cloud Hosting
    Shared Web Hosting
    Consulting Services

  3. #3
    LMStone's Avatar
    LMStone is offline Moderator
    Join Date
    Sep 2006
    Location
    477 Congress Street | Portland, ME 04101
    Posts
    1,374
    Rep Power
    10

    Default

    Postfix is an MTA only. We have built Postfix boxes to handle 20,000 emails per hour with 1.5GB RAM.

    Zimbra is a full groupware suite with a searchable mailstore and lots of other software bits. So, a Postfix-only to Zimbra comparison is not an apples-to-apples comparison IMHO.

    Mark

  4. #4
    jruiseco is offline Intermediate Member
    Join Date
    May 2011
    Location
    Colombia
    Posts
    22
    Rep Power
    4

    Default Logs

    I'd like to focus more on my log errors. Can anybody explain to me what's going on in the log errors above? Does the configuration sound right? How can I speed up webmail? Is there any howto? IMAP and POP are ok, but I do get these errors in my logs regarding I/O errors.

    Thanks,

    JR

  5. #5
    Krishopper is offline Dedicated Member
    Join Date
    Dec 2006
    Location
    Minneapolis MN
    Posts
    777
    Rep Power
    9

    Default

    What email client(s) are your users using? In particular, the user referenced in the error log above?
    01 Networks, LLC / Cybernetik.net
    Zimbra NE and OSS Cloud Hosting
    Shared Web Hosting
    Consulting Services

  6. #6
    jruiseco is offline Intermediate Member
    Join Date
    May 2011
    Location
    Colombia
    Posts
    22
    Rep Power
    4

    Default

    The clients are most likely outlook, since ZD uses soap, not IMAP (right?).

    Regarding swapping, the box has used up 50M of swap space, but when I restart the server it's not released so I'm inclined to think zimbra itsself isn't swapping. dstat shows no paging activity during daytime operation.

    Yes the box is a virtual box, I'll post my benchmark results in a bit for comparison's sake.

    Another issue also is that I come from a perl/c/apache background so essentially we managed more capacity by using more apache processes. Is there anything tuneable in the webmail interface to make it serve faster? I hear a lot about caches, etc but I have no clue how to change the settings. I've never managed a java server before. Any recommendationas as to where I should look?

    JR
    Last edited by jruiseco; 05-03-2011 at 12:41 PM. Reason: Forgot to answer question above.

  7. #7
    LMStone's Avatar
    LMStone is offline Moderator
    Join Date
    Sep 2006
    Location
    477 Congress Street | Portland, ME 04101
    Posts
    1,374
    Rep Power
    10

    Default

    Quote Originally Posted by jruiseco View Post
    The clients are most likely outlook, since ZD uses soap, not IMAP (right?).

    Regarding swapping, the box has used up 50M of swap space, but when I restart the server it's not released so I'm inclined to think zimbra itsself isn't swapping. dstat shows no paging activity during daytime operation.

    Yes the box is a virtual box, I'll post my benchmark results in a bit for comparison's sake.

    Another issue also is that I come from a perl/c/apache background so essentially we managed more capacity by using more apache processes. Is there anything tuneable in the webmail interface to make it serve faster? I hear a lot about caches, etc but I have no clue how to change the settings. I've never managed a java server before. Any recommendationas as to where I should look?

    JR
    Can you tell us something of your virtualization environment? Physical host, virtualization software, other virtual machines running on it, network connectivity, etc.

    Speeding up the web interface generally takes RAM. But 4GB for 50 users shouldn't be an issue at all.

    Did you by any chance increase/decrease the amount of RAM you allocated to the Zimbra VM after you did the Zimbra install?

    Can you also as the zimbra user (after becoming root do "su - zimbra" and apologies if I am telling you something you already know) run the following three commands and report the results here please?

    Code:
    cat /opt/zimbra/conf/my.cnf | grep innodb_
    
    zmlocalconfig | grep java_heap
    
    vmstat -n 1  (about 20 lines of output while the server is busy should do)

  8. #8
    jruiseco is offline Intermediate Member
    Join Date
    May 2011
    Location
    Colombia
    Posts
    22
    Rep Power
    4

    Default

    Sure, first I'll describe the hardware, just so we can be clear it's not just any "virtual box":

    RackForce cloud box with 2 vcores (they call them VPUs), the benchmark showed that it can can serve 8813.54 requests per second in comparison to 9840.56 requests from my AMD 3600MHz box.

    Their drive is a SAN disk of 150GB, benchmark averaged 192.42 MB/s on write for 37 minutes with 12 clients (phoronix dbench) and I saw bursts of up to 310MB/s. In comparison my Athlon II X2 running at 3600MHz did 62.79 MB/s writing to an lvm striped volume, running on 2 WD1600AAJS SATA drives (not slow).

    Now to the config part:

    #cat /opt/zimbra/conf/my.cnf | grep innodb_

    [root@crm ~]# cat /opt/zimbra/conf/my.cnf |grep innodb_
    innodb_buffer_pool_size = 632685772
    innodb_log_file_size = 104857600
    innodb_log_buffer_size = 8388608
    innodb_file_per_table
    innodb_open_files = 2710
    innodb_max_dirty_pages_pct = 10
    innodb_flush_method = O_DIRECT
    innodb_flush_log_at_trx_commit = 0

    #zmlocalconfig | grep java_heap
    mailboxd_java_heap_memory_percent = 20
    mailboxd_java_heap_new_size_percent = 25

    #vmstat -n 1
    procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu------
    r b swpd free buff cache si so bi bo in cs us sy id wa st
    1 0 49468 601344 309456 1254812 0 0 2 67 2 2 16 1 82 0 0
    0 0 49468 610684 309456 1254780 0 0 0 568 1036 1782 28 6 63 2 0
    0 0 49468 610684 309456 1254776 0 0 0 404 1032 325 0 0 100 0 0
    1 0 49468 610700 309456 1254776 0 0 0 0 1023 305 0 0 100 0 0
    0 0 49468 610700 309456 1254776 0 0 0 24 1018 340 1 0 100 0 0
    0 0 49468 610700 309456 1254776 0 0 0 16 1019 311 0 0 100 0 0
    0 0 49468 610700 309456 1254776 0 0 0 44 1026 340 0 0 100 1 0
    0 0 49468 610824 309456 1254776 0 0 0 0 1012 307 0 0 100 0 0
    0 0 49468 610824 309456 1254776 0 0 0 0 1014 310 0 0 100 0 0
    0 0 49468 610816 309456 1254776 0 0 0 64 1029 357 0 0 99 0 0
    0 0 49468 610824 309456 1254776 0 0 0 0 1018 335 0 0 100 0 0
    0 0 49468 610824 309456 1254776 0 0 0 420 1050 414 1 0 98 2 0
    0 0 49468 610824 309456 1254776 0 0 0 20 1012 312 0 0 100 0 0
    0 0 49468 612080 309456 1254776 0 0 0 0 1018 315 0 0 100 0 0
    0 0 49468 612080 309456 1254776 0 0 0 0 1012 328 0 0 100 0 0
    0 0 49468 612080 309456 1254776 0 0 0 752 1119 327 0 0 100 1 0
    0 0 49468 612080 309456 1254776 0 0 0 0 1012 339 1 0 100 0 0
    0 0 49468 611956 309456 1254780 0 0 4 32 1024 362 0 1 100 0 0
    0 0 49468 612328 309456 1254780 0 0 0 256 1021 440 2 0 97 1 0

  9. #9
    LMStone's Avatar
    LMStone is offline Moderator
    Join Date
    Sep 2006
    Location
    477 Congress Street | Portland, ME 04101
    Posts
    1,374
    Rep Power
    10

    Default

    I'm fine that RackForce is using good hardware, networking and VMware Enterprise Plus.

    What you can't tell is the extent to which RackForce may be over-subscribing the physical hardware. Say the compute head on which your VM is domiciled has 2x12 core CPUs and 10 4-core virtual machines. That's a lot of context switching, especially if there are other busy virtual machines on your physical host.

    If you don't mind providing a test user account on your Zimbra server, I'd be happy to log in to the web interface and see how snappy it is or is not.

    One step at a time...

    Hope that helps,
    Mark

  10. #10
    jruiseco is offline Intermediate Member
    Join Date
    May 2011
    Location
    Colombia
    Posts
    22
    Rep Power
    4

    Default

    Yeah, why not, speed is relative, right?

    url is Zimbra Collaboration Suite Log In user is lmstone and password is LMStone456& you'll have to change that at first login.

    So nothing blaringly bad in my setup this far? Let me ask you a question, what's the difference between zmlocalconfig -e zimbra_session_limit_imap=500 and changing the max threads from the GUI?

    JR

Page 1 of 2 12 LastLast

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Similar Threads

  1. fatal: parameter "smtpd_recipient_restrictions"
    By Robin in forum Administrators
    Replies: 8
    Last Post: 12-22-2010, 05:48 AM
  2. need advice on configuring zimbra to work with fax server
    By pheonix1t in forum Administrators
    Replies: 0
    Last Post: 07-11-2007, 07:46 PM
  3. Can't start Zimbra!
    By zibra in forum Administrators
    Replies: 5
    Last Post: 03-22-2007, 11:34 AM
  4. Post instsallation problems
    By Assaf in forum Installation
    Replies: 14
    Last Post: 01-29-2007, 11:38 AM
  5. Zimbra server crashed
    By goetzi in forum Administrators
    Replies: 6
    Last Post: 03-25-2006, 01:00 PM

Posting Permissions

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