Is ClusterControl normally this slow?

Comments

12 comments

  • Avatar
    Johan

    Hi Gabe,

    No this does not look right.

    Can you login to the controller and do (as sudo/root):

    free -m

    vmstat 1 30

    # vmstat will run for 30 seconds

    top -d1

    #press capital M while the top is running, copy/paste as much as you can here.

    the please do, on the controller:

    mysql -ucmon -p -h127.0.0.1 cmon -e "show processlist"

    Please send here the output of the commands.

    If you do:

    apachectl restart

    Does it help?

     

    * What version of ClusterControl are you using?

    * What version of MySQL do you have?

    BR 

    johan

     

    0
    Comment actions Permalink
  • Avatar
    Gabe

    Johan,

    Thanks for the reply.  I have collected the information you have requested.  I sanitized the hostname in one of the files.  Restarting Apache does not resolve the issue.

    Thanks,

    Gabe

    0
    Comment actions Permalink
  • Avatar
    Gabe

    I forgot, the clustercontrol I am using was downloaded off your site yesterday. 

    clustercontrol-cmonapi-1.2.8-51.x86_64
    clustercontrol-controller-1.2.8-490.x86_64
    clustercontrol-1.2.8-54.x86_64

    I am running the controller database on Percona Server 5.6.

    0
    Comment actions Permalink
  • Avatar
    Johan

    Hi,

    In the output from top (top.txt) i could not see the mysql server. 

    Also, i could not find any httpd/apache process.

    Is the cmon controller and apache and mysqld (for the cmon database) running on the same server?

    Can you create a new 'top' if they are indeed running on the same server, or create the top from the server where the apache + mysqld are running. 

    Besides that, there is no load on the server, and swap isn't used.

    Can you do like this:

    When the UI is rendering the page (and nothing really happens), could you do,

    mysql -ucmon -p -h127.0.0.1 cmon -e "show full processlist; select sleep(1); show full processlist; select sleep(1); show full processlist" 

    If nothing there, i am thinking network issues in the area of  DNS,  Firewall,  dropped packages.

    Run on the controller (or where the apache is running):

    time curl "http://127.0.0.1/cmonapi/nodes/all.json?clusterid=1"

    time curl "http://127.0.0.1/cmonapi/clusters/dbload.json?clusterid=1"

    What do you get? (if you get a lot of data back you can discard that, i am not interested in the actual data returned, more that it works).

    Best regards

    Johan

    0
    Comment actions Permalink
  • Avatar
    Gabe

    Johan,

    I've attached a complete top output (using batch mode) and also the mysql process list while the UI is stuck.  I ran your two curl commands, the first completed in .162s and the second in .154s.

    Is it possible that its an issue with the nodes being added with dns names?  Is it possible to reconfigure it with IP addresses for the hosts, or should I just drop the whole cluster and re-add?

    Gabe

    0
    Comment actions Permalink
  • Avatar
    Johan

    Hi,

    I don't know yet if it is a DNS isssue, but could be. I would need to check with a colleague but it could take a while (he is not online now).

    Did you use "add existing cluster" or did you create using our Galera configurator? 

    BR
    johan

    0
    Comment actions Permalink
  • Avatar
    Gabe

    I added an existing cluster as our boxes have been up and running for a long time.

    Gabe

    0
    Comment actions Permalink
  • Avatar
    Ashraf Sharif

    Hi Gabe,

    Did you configure PTR record for all IP addresses in the cluster (ClusterControl and PXC nodes)? You can verify with following commands:

    dig ptr <ip address of CC or PXC nodes>

    Ensure you get an "ANSWER SECTION" or just use the simple host command:

    host <IP address of CC or PXC nodes>

    Ensure you don't get a "not found: 3(NXDOMAIN)"

     

    If no, then you may need to use IP address instead. From the ClusterControl UI > Database Clusters > just press the Delete > tick Completely remove it from Controller. Then, add following line into /etc/my.cnf:

    skip-name-resolve

     

    Restart MySQL and then re-add again the cluster using IP address from ClusterControl UI.

     

    Regards,

    Ashraf

     

    0
    Comment actions Permalink
  • Avatar
    Gabe

    Ashraf,

    All database nodes and the cluster control node have DNS entries.  "host XXX" resolves all the IPs correctly.  Any other ideas?

    Gabe

    0
    Comment actions Permalink
  • Avatar
    Jeremy Hutchings

    Hi Gabe / all,

     

    Did you ever figure this out ? I've just installed CC with a 3 node cluster, all ssh keys and connectivity working fine. Then go to the UI and add the cluster, and the job runs fine. Then the UI just hangs with "Loading".

    I notice a lot of tables like mysql_server and node_state are all empty.

    The cluster is called default_name()

    And then there is CONTROLLER:  and GALERA NODES: which I'd expect to have a 1 and 3 after, but it's blanks.

    I thought CC would of detected and added the nodes opposed to me having to add them all individually ?

    0
    Comment actions Permalink
  • Avatar
    Ashraf Sharif

    Hi Jeremy,

    I suggest you to create a support ticket and provide us with error report as described in this page: http://support.severalnines.com/entries/21091498-Error-reporting-

     

    Regards,

    Ashraf

    0
    Comment actions Permalink
  • Avatar
    Jeremy Hutchings

    Hi Asharf,

    I think I have it working now, I just removed and added the cluster a few times and it found it on the 4th go.

    Thanks

    K

    0
    Comment actions Permalink

Please sign in to leave a comment.

Powered by Zendesk