Could you share the glusterd log file and the scripts which were triggered? ~Atin On 03/18/2015 04:14 PM, F?lix de Lelelis wrote:> Hi, > > Sorry, I have not explained well. I don't mean that I have lunch this > comand simultaneously, I am using zabbix and I have defined diferent time > intervals for this but comes a point at wich I can't luch anythig command > over the cluster, such as "gluster volume status". Is there any way around > this? > > Thanks > > > > 2015-03-18 11:36 GMT+01:00 Atin Mukherjee <amukherj at redhat.com>: > >> >> >> On 03/18/2015 03:24 PM, F?lix de Lelelis wrote: >>> Hi, >>> >>> I have a problem with glusterfs 3.6. I am monitoring it with scripts that >>> lunch "gluster volume status VONAME detail" and "gluster volume profile >>> VOLNAME info". When this scripts are running about 1-2 hours, with check >>> every 1 minute, gluster is blocked and the node generates "Another >>> transaction is in progress". The only way to restore the situation is >>> restart glusterd. >>> >>> Can you help me please. >> This is expected. Concurrent transactions on a same volumes are not >> allowed. There is a cluster wide volume lock maintained by glusterd. In >> this case since both these commands are for VOLNAME, any one of two >> would throw an error saying "Another transaction is in progress". If you >> ensure that there is a time gap between these two commands in your >> script, you can get rid of this situation. >> >> ~Atin >>> >>> >>> Thanks >>> >>> >>> >>> _______________________________________________ >>> Gluster-users mailing list >>> Gluster-users at gluster.org >>> http://www.gluster.org/mailman/listinfo/gluster-users >>> >> >> -- >> ~Atin >> >-- ~Atin
Hi Atin, I send you the log and scripts that I am using. Wait a response. Thanks. 2015-03-18 15:21 GMT+01:00 Atin Mukherjee <amukherj at redhat.com>:> Could you share the glusterd log file and the scripts which were triggered? > > ~Atin > > On 03/18/2015 04:14 PM, F?lix de Lelelis wrote: > > Hi, > > > > Sorry, I have not explained well. I don't mean that I have lunch this > > comand simultaneously, I am using zabbix and I have defined diferent time > > intervals for this but comes a point at wich I can't luch anythig command > > over the cluster, such as "gluster volume status". Is there any way > around > > this? > > > > Thanks > > > > > > > > 2015-03-18 11:36 GMT+01:00 Atin Mukherjee <amukherj at redhat.com>: > > > >> > >> > >> On 03/18/2015 03:24 PM, F?lix de Lelelis wrote: > >>> Hi, > >>> > >>> I have a problem with glusterfs 3.6. I am monitoring it with scripts > that > >>> lunch "gluster volume status VONAME detail" and "gluster volume profile > >>> VOLNAME info". When this scripts are running about 1-2 hours, with > check > >>> every 1 minute, gluster is blocked and the node generates "Another > >>> transaction is in progress". The only way to restore the situation is > >>> restart glusterd. > >>> > >>> Can you help me please. > >> This is expected. Concurrent transactions on a same volumes are not > >> allowed. There is a cluster wide volume lock maintained by glusterd. In > >> this case since both these commands are for VOLNAME, any one of two > >> would throw an error saying "Another transaction is in progress". If you > >> ensure that there is a time gap between these two commands in your > >> script, you can get rid of this situation. > >> > >> ~Atin > >>> > >>> > >>> Thanks > >>> > >>> > >>> > >>> _______________________________________________ > >>> Gluster-users mailing list > >>> Gluster-users at gluster.org > >>> http://www.gluster.org/mailman/listinfo/gluster-users > >>> > >> > >> -- > >> ~Atin > >> > > > > -- > ~Atin >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150320/b1fe8b1e/attachment-0001.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: log_script_gfs.zip Type: application/zip Size: 6508018 bytes Desc: not available URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150320/b1fe8b1e/attachment-0001.zip>
I could see you have executed profile commands in background in a 1 sec time interval which means if profile command takes more than a second there will be concurrent transactions coming to glusterd at a given point of time which could lead to "Another transaction is in progress" situation. However you shouldn't end up with a stale lock by any means. Glusterd log which you have provided is log rotated and doesn't have the entry where the transaction failed because of the another transaction was going through. Do you also have a single node cluster, if not I would require glusterd logs for all the nodes. ~Atin On 03/20/2015 12:43 PM, F?lix de Lelelis wrote:> Hi Atin, > > I send you the log and scripts that I am using. > Wait a response. > > Thanks. > > 2015-03-18 15:21 GMT+01:00 Atin Mukherjee <amukherj at redhat.com>: > >> Could you share the glusterd log file and the scripts which were triggered? >> >> ~Atin >> >> On 03/18/2015 04:14 PM, F?lix de Lelelis wrote: >>> Hi, >>> >>> Sorry, I have not explained well. I don't mean that I have lunch this >>> comand simultaneously, I am using zabbix and I have defined diferent time >>> intervals for this but comes a point at wich I can't luch anythig command >>> over the cluster, such as "gluster volume status". Is there any way >> around >>> this? >>> >>> Thanks >>> >>> >>> >>> 2015-03-18 11:36 GMT+01:00 Atin Mukherjee <amukherj at redhat.com>: >>> >>>> >>>> >>>> On 03/18/2015 03:24 PM, F?lix de Lelelis wrote: >>>>> Hi, >>>>> >>>>> I have a problem with glusterfs 3.6. I am monitoring it with scripts >> that >>>>> lunch "gluster volume status VONAME detail" and "gluster volume profile >>>>> VOLNAME info". When this scripts are running about 1-2 hours, with >> check >>>>> every 1 minute, gluster is blocked and the node generates "Another >>>>> transaction is in progress". The only way to restore the situation is >>>>> restart glusterd. >>>>> >>>>> Can you help me please. >>>> This is expected. Concurrent transactions on a same volumes are not >>>> allowed. There is a cluster wide volume lock maintained by glusterd. In >>>> this case since both these commands are for VOLNAME, any one of two >>>> would throw an error saying "Another transaction is in progress". If you >>>> ensure that there is a time gap between these two commands in your >>>> script, you can get rid of this situation. >>>> >>>> ~Atin >>>>> >>>>> >>>>> Thanks >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> Gluster-users mailing list >>>>> Gluster-users at gluster.org >>>>> http://www.gluster.org/mailman/listinfo/gluster-users >>>>> >>>> >>>> -- >>>> ~Atin >>>> >>> >> >> -- >> ~Atin >> >-- ~Atin