Interesting table Karan!, Could you please tell us how you did the benchmark? fio or iozone orsimilar? thanks Arman. On Wed, Sep 27, 2017 at 1:20 PM, Karan Sandha <ksandha at redhat.com> wrote:> Hi Collin, > > During our arbiter latency testing for completion of ops we found the > below results:- an arbiter node in another data centre and both the data > bricks in the same data centre, > > 1) File-size 1 KB (10000 files ) > 2) mkdir > > > Latency > > 5ms > > 10ms > > 20ms > > 50ms > > 100ms > > 200ms > > Ops > > Create > > 755 secs > > 1410 secs > > 2717 secs > > 5874 secs > > 12908 sec > > 26113 sec > > Mkdir > > 922 secs > > 1725 secs > > 3325 secs > > 8127 secs > > 16160 sec > > 30079 sec > > > Thanks & Regards > > On Mon, Sep 25, 2017 at 5:40 AM, Colin Coe <colin.coe at gmail.com> wrote: > >> Hi all >> >> I've googled but can't find an answer to my question. >> >> I have two data centers. Currently, I have a replica (count of 2 plus >> arbiter) in one data center but is used by both. >> >> I want to change this to be a distributed replica across the two data >> centers. >> >> There is a 20Mbps pipe and approx 22 ms latency. Is this sufficient? >> >> I really don't want to do the geo-replication in its current form. >> >> Thanks >> >> CC >> >> _______________________________________________ >> Gluster-users mailing list >> Gluster-users at gluster.org >> http://lists.gluster.org/mailman/listinfo/gluster-users >> > > > > -- > > KARAN SANDHA > > ASSOCIATE QUALITY ENGINEER > > Red Hat Bangalore <https://www.redhat.com/> > > ksandha at redhat.com M: 9888009555 IM: Karan on @irc > <https://red.ht/sig> > TRIED. TESTED. TRUSTED. <https://redhat.com/trusted> > @redhatnews <https://twitter.com/redhatnews> Red Hat > <https://www.linkedin.com/company/red-hat> Red Hat > <https://www.facebook.com/RedHatInc> > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://lists.gluster.org/mailman/listinfo/gluster-users >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170928/92624e2c/attachment.html>
It was simple emulation of network packets on the port of the server node using tc tool tc qdisc add dev <port> root netem delay <time>ms. The files were created using dd tool (in-built in linux) and mkdir. Post the IO's we verified with no pending heals. Thanks & Regards On Thu, Sep 28, 2017 at 2:06 PM, Arman Khalatyan <arm2arm at gmail.com> wrote:> Interesting table Karan!, > Could you please tell us how you did the benchmark? fio or iozone > orsimilar? > > thanks > Arman. > > On Wed, Sep 27, 2017 at 1:20 PM, Karan Sandha <ksandha at redhat.com> wrote: > >> Hi Collin, >> >> During our arbiter latency testing for completion of ops we found the >> below results:- an arbiter node in another data centre and both the data >> bricks in the same data centre, >> >> 1) File-size 1 KB (10000 files ) >> 2) mkdir >> >> >> Latency >> >> 5ms >> >> 10ms >> >> 20ms >> >> 50ms >> >> 100ms >> >> 200ms >> >> Ops >> >> Create >> >> 755 secs >> >> 1410 secs >> >> 2717 secs >> >> 5874 secs >> >> 12908 sec >> >> 26113 sec >> >> Mkdir >> >> 922 secs >> >> 1725 secs >> >> 3325 secs >> >> 8127 secs >> >> 16160 sec >> >> 30079 sec >> >> >> Thanks & Regards >> >> On Mon, Sep 25, 2017 at 5:40 AM, Colin Coe <colin.coe at gmail.com> wrote: >> >>> Hi all >>> >>> I've googled but can't find an answer to my question. >>> >>> I have two data centers. Currently, I have a replica (count of 2 plus >>> arbiter) in one data center but is used by both. >>> >>> I want to change this to be a distributed replica across the two data >>> centers. >>> >>> There is a 20Mbps pipe and approx 22 ms latency. Is this sufficient? >>> >>> I really don't want to do the geo-replication in its current form. >>> >>> Thanks >>> >>> CC >>> >>> _______________________________________________ >>> Gluster-users mailing list >>> Gluster-users at gluster.org >>> http://lists.gluster.org/mailman/listinfo/gluster-users >>> >> >> >> >> -- >> >> KARAN SANDHA >> >> ASSOCIATE QUALITY ENGINEER >> >> Red Hat Bangalore <https://www.redhat.com/> >> >> ksandha at redhat.com M: 9888009555 IM: Karan on @irc >> <https://red.ht/sig> >> TRIED. TESTED. TRUSTED. <https://redhat.com/trusted> >> @redhatnews <https://twitter.com/redhatnews> Red Hat >> <https://www.linkedin.com/company/red-hat> Red Hat >> <https://www.facebook.com/RedHatInc> >> >> _______________________________________________ >> Gluster-users mailing list >> Gluster-users at gluster.org >> http://lists.gluster.org/mailman/listinfo/gluster-users >> > >-- KARAN SANDHA ASSOCIATE QUALITY ENGINEER Red Hat Bangalore <https://www.redhat.com/> ksandha at redhat.com M: 9888009555 IM: Karan on @irc <https://red.ht/sig> TRIED. TESTED. TRUSTED. <https://redhat.com/trusted> @redhatnews <https://twitter.com/redhatnews> Red Hat <https://www.linkedin.com/company/red-hat> Red Hat <https://www.facebook.com/RedHatInc> -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170929/5131988e/attachment.html>
Apologies for the late reply. Further to this, if my Linux clients are connecting uing glusterfs-fuse and I have my volumes defined like this: dc1srv1:/gv_fileshare dc2srv1:/gv_fileshare dc1srv2:/gv_fileshare dc2srv2:/gv_fileshare (replica 2) How do I ensure that clients in dc1 prefer dc1srv1 and dc1srv2 while clients in dc2 prefer the dc2 servers? Is it simply a matter of ordering in /etc/fstab? Thanks On Fri, Sep 29, 2017 at 2:29 PM, Karan Sandha <ksandha at redhat.com> wrote:> It was simple emulation of network packets on the port of the server node > using tc tool tc qdisc add dev <port> root netem delay <time>ms. The > files were created using dd tool (in-built in linux) and mkdir. Post the > IO's we verified with no pending heals. > > > Thanks & Regards > > > On Thu, Sep 28, 2017 at 2:06 PM, Arman Khalatyan <arm2arm at gmail.com> > wrote: > >> Interesting table Karan!, >> Could you please tell us how you did the benchmark? fio or iozone >> orsimilar? >> >> thanks >> Arman. >> >> On Wed, Sep 27, 2017 at 1:20 PM, Karan Sandha <ksandha at redhat.com> wrote: >> >>> Hi Collin, >>> >>> During our arbiter latency testing for completion of ops we found the >>> below results:- an arbiter node in another data centre and both the data >>> bricks in the same data centre, >>> >>> 1) File-size 1 KB (10000 files ) >>> 2) mkdir >>> >>> >>> Latency >>> >>> 5ms >>> >>> 10ms >>> >>> 20ms >>> >>> 50ms >>> >>> 100ms >>> >>> 200ms >>> >>> Ops >>> >>> Create >>> >>> 755 secs >>> >>> 1410 secs >>> >>> 2717 secs >>> >>> 5874 secs >>> >>> 12908 sec >>> >>> 26113 sec >>> >>> Mkdir >>> >>> 922 secs >>> >>> 1725 secs >>> >>> 3325 secs >>> >>> 8127 secs >>> >>> 16160 sec >>> >>> 30079 sec >>> >>> >>> Thanks & Regards >>> >>> On Mon, Sep 25, 2017 at 5:40 AM, Colin Coe <colin.coe at gmail.com> wrote: >>> >>>> Hi all >>>> >>>> I've googled but can't find an answer to my question. >>>> >>>> I have two data centers. Currently, I have a replica (count of 2 plus >>>> arbiter) in one data center but is used by both. >>>> >>>> I want to change this to be a distributed replica across the two data >>>> centers. >>>> >>>> There is a 20Mbps pipe and approx 22 ms latency. Is this sufficient? >>>> >>>> I really don't want to do the geo-replication in its current form. >>>> >>>> Thanks >>>> >>>> CC >>>> >>>> _______________________________________________ >>>> Gluster-users mailing list >>>> Gluster-users at gluster.org >>>> http://lists.gluster.org/mailman/listinfo/gluster-users >>>> >>> >>> >>> >>> -- >>> >>> KARAN SANDHA >>> >>> ASSOCIATE QUALITY ENGINEER >>> >>> Red Hat Bangalore <https://www.redhat.com/> >>> >>> ksandha at redhat.com M: 9888009555 IM: Karan on @irc >>> <https://red.ht/sig> >>> TRIED. TESTED. TRUSTED. <https://redhat.com/trusted> >>> @redhatnews <https://twitter.com/redhatnews> Red Hat >>> <https://www.linkedin.com/company/red-hat> Red Hat >>> <https://www.facebook.com/RedHatInc> >>> >>> _______________________________________________ >>> Gluster-users mailing list >>> Gluster-users at gluster.org >>> http://lists.gluster.org/mailman/listinfo/gluster-users >>> >> >> > > > -- > > KARAN SANDHA > > ASSOCIATE QUALITY ENGINEER > > Red Hat Bangalore <https://www.redhat.com/> > > ksandha at redhat.com M: 9888009555 IM: Karan on @irc > <https://red.ht/sig> > TRIED. TESTED. TRUSTED. <https://redhat.com/trusted> > @redhatnews <https://twitter.com/redhatnews> Red Hat > <https://www.linkedin.com/company/red-hat> Red Hat > <https://www.facebook.com/RedHatInc> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20171013/7caf270d/attachment.html>