Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-16536

memcached fd leak caused rebalance to fail

    XMLWordPrintable

Details

    • Bug
    • Resolution: Cannot Reproduce
    • Major
    • None
    • 3.1.0
    • memcached
    • Security Level: Public
    • Untriaged
    • Centos 64-bit
    • Unknown

    Description

      Rebalance kept failing on one node.

      There are many memcached connections which exceed the max limits 30,000.

      The memcached logs also show the same:

      Thu Aug 13 18:28:49.832053 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832077 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832106 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832128 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832146 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832171 GMT 3: Too many open connections
      Thu Aug 13 18:28:49.832186 GMT 3: Too many open connections
      

      Also from lsof output, there are many below entries:

      memcached 5541 couchbase 1422u sock 0,6 0t0 749862 can't identify protocol
      memcached 5541 couchbase 1423u sock 0,6 0t0 773067 can't identify protocol
      memcached 5541 couchbase 1424u sock 0,6 0t0 629256 can't identify protocol
      memcached 5541 couchbase 1425u sock 0,6 0t0 826786 can't identify protocol
      memcached 5541 couchbase 1426u sock 0,6 0t0 653344 can't identify protocol
      memcached 5541 couchbase 1427u sock 0,6 0t0 677622 can't identify protocol
      memcached 5541 couchbase 1428u sock 0,6 0t0 151754 can't identify protocol
      memcached 5541 couchbase 1429u sock 0,6 0t0 637362 can't identify protocol
      memcached 5541 couchbase 1430u sock 0,6 0t0 53756963 can't identify protocol
      memcached 5541 couchbase 1431u sock 0,6 0t0 269680 can't identify protocol
      memcached 5541 couchbase 1432u sock 0,6 0t0 490188 can't identify protocol
      

      After some research, those "can't identify protocol" messages indicates that the system reach the open file limit for couchbase user whose ulimit of open file is set to 40960.

      I think there is a possible FD leak in memcached.

      Attachments

        No reviews matched the request. Check your Options in the drop-down menu of this sections header.

        Activity

          People

            james.mauss James Mauss (Inactive)
            larry Larry Liu [X] (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Gerrit Reviews

                There are no open Gerrit changes

                PagerDuty