Hi Pradeep, For changing the snitch you have decommission and add the node with new switch and updated properties files.
Thanks, > On Aug 26, 2018, at 2:15 PM, Joshua Galbraith > <jgalbra...@newrelic.com.INVALID> wrote: > > Pradeep, > > Here are some related tickets that may also be helpful in understanding the > current behavior of these options. > > * https://issues.apache.org/jira/browse/CASSANDRA-5897 > <https://issues.apache.org/jira/browse/CASSANDRA-5897> > * https://issues.apache.org/jira/browse/CASSANDRA-9474 > <https://issues.apache.org/jira/browse/CASSANDRA-9474> > * https://issues.apache.org/jira/browse/CASSANDRA-10243 > <https://issues.apache.org/jira/browse/CASSANDRA-10243> > * https://issues.apache.org/jira/browse/CASSANDRA-10242 > <https://issues.apache.org/jira/browse/CASSANDRA-10242> > > On Sun, Aug 26, 2018 at 1:20 PM, Joshua Galbraith <jgalbra...@newrelic.com > <mailto:jgalbra...@newrelic.com>> wrote: > Pradeep, > > That being said, I haven't experimented with -Dcassandra.ignore_dc=true > -Dcassandra.ignore_rack=true before. > > The description here may be helpful: > https://github.com/apache/cassandra/blob/trunk/NEWS.txt#L685-L693 > <https://github.com/apache/cassandra/blob/trunk/NEWS.txt#L685-L693> > > I would spin up a small test cluster with data you don't care about and > verify that your above assumptions are correct there first. > > On Sun, Aug 26, 2018 at 1:09 PM, Joshua Galbraith <jgalbra...@newrelic.com > <mailto:jgalbra...@newrelic.com>> wrote: > Pradeep. > > Right, so from that documentation is sounds like you actually have to stop > all nodes in the cluster at once and bring them back up one at a time. A > rolling restart won't work here. > > On Sun, Aug 26, 2018 at 11:46 AM, Pradeep Chhetri <prad...@stashaway.com > <mailto:prad...@stashaway.com>> wrote: > Hi Joshua, > > Thank you for the reply. Sorry i forgot to mention that I already went > through that documentation. There are few missing things regarding which I > have few questions: > > 1) One thing which isn't mentioned there is that cassandra fails to restart > when we change the datacenter name or rack name of a node. So whether should > i first rolling restart cassandra with flag "-Dcassandra.ignore_dc=true > -Dcassandra.ignore_rack=true", then run sequential repair and then cleanup > and then rolling restart cassandra without that flag. > > 2) Should i not allow any read/write operation from applications during the > time when sequential repair is running. > > Regards, > Pradeep > > On Mon, Aug 27, 2018 at 12:19 AM, Joshua Galbraith > <jgalbra...@newrelic.com.invalid <mailto:jgalbra...@newrelic.com.invalid>> > wrote: > Pradeep, it sounds like what you're proposing counts as a topology change > because you are changing the datacenter name and rack name. > > Please refer to the documentation here about what to do in that situation: > https://docs.datastax.com/en/cassandra/3.0/cassandra/operations/opsSwitchSnitch.html > > <https://docs.datastax.com/en/cassandra/3.0/cassandra/operations/opsSwitchSnitch.html> > > In particular: > > Simply altering the snitch and replication to move some nodes to a new > datacenter will result in data being replicated incorrectly. > > Topology changes may occur when the replicas are placed in different places > by the new snitch. Specifically, the replication strategy places the replicas > based on the information provided by the new snitch. > > If the topology of the network has changed, but no datacenters are added: > a. Shut down all the nodes, then restart them. > b. Run a sequential repair and nodetool cleanup on each node. > > On Sun, Aug 26, 2018 at 11:14 AM, Pradeep Chhetri <prad...@stashaway.com > <mailto:prad...@stashaway.com>> wrote: > Hello everyone, > > Since i didn't hear from anyone, just want to describe my question again: > > Am i correct in understanding that i need to do following steps to migrate > data from SimpleSnitch to GPFS changing datacenter name and rack name to AWS > region and Availability zone respectively > > 1) Update the rack and datacenter fields in cassandra-rackdc.properties file > and rolling restart cassandra with this flag "-Dcassandra.ignore_dc=true > -Dcassandra.ignore_rack=true" > > 2) Run nodetool repair --sequential and nodetool cleanup. > > 3) Rolling restart cassandra removing the flag "-Dcassandra.ignore_dc=true > -Dcassandra.ignore_rack=true" > > Regards, > Pradeep > > On Thu, Aug 23, 2018 at 10:53 PM, Pradeep Chhetri <prad...@stashaway.com > <mailto:prad...@stashaway.com>> wrote: > Hello, > > I am currently running a 3.11.2 cluster in SimpleSnitch hence the datacenter > is datacenter1 and rack is rack1 for all nodes on AWS. I want to switch to > GPFS by changing the rack name to the availability-zone name and datacenter > name to region name. > > When I try to restart individual nodes by changing those values, it failed to > start throwing the error about dc and rack name mismatch but gives me an > option to set ignore_dc and ignore_rack to true to bypass it. > > I am not sure if it is safe to set those two flags to true and if there is > any drawback now or in future when i add a new datacenter to the cluster. I > went through the documentation on Switching Snitches but didn't get much > explanation. > > Regards, > Pradeep > > > > > > > > > > > > -- > Joshua Galbraith | Lead Software Engineer | New Relic > > > > > -- > Joshua Galbraith | Lead Software Engineer | New Relic > > > > -- > Joshua Galbraith | Lead Software Engineer | New Relic > > > > -- > Joshua Galbraith | Lead Software Engineer | New Relic
