Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cassandra.apache.org
Received-SPF: pass (nike.apache.org: local policy)
Content-Type: text/plain; charset=iso-8859-1
Mime-Version: 1.0 (Mac OS X Mail 6.2 \(1499\))
Subject: Re: Ec2Snitch to Ec2MultiRegionSnitch
From: aaron morton <aaron@thelastpickle.com>
In-Reply-To: 
 <CA+VSrLqZQYFs_F6QQjKmJuMxuU1e_cJF1ap_kWs5UccpxyNmeQ@mail.gmail.com>
Date: Wed, 24 Apr 2013 15:33:36 +1200
Content-Transfer-Encoding: quoted-printable
Message-Id: <AC160CF4-15A9-4CD2-B94F-E80979A5507E@thelastpickle.com>
References: 
 <CA+VSrLovBjjzhrtQhgXxTHA5kELfpWnAFHbKFEmeLU-HGUm=1A@mail.gmail.com>
 <CAL7B-cDohXD8pWQ_FVGWXjPrHVGjkr84ORUsN++3_fbbPcfrug@mail.gmail.com>
 <CA+VSrLqZQYFs_F6QQjKmJuMxuU1e_cJF1ap_kWs5UccpxyNmeQ@mail.gmail.com>
To: user@cassandra.apache.org

> You are advising me to test it, what would be a good way of testing it =
(I can use AWS EC2 instances if needed) ?
If you are only using one Available Zone per region then you have only =
one rack per DC and the NetworkTopologyStrategy will do the right thing.=20=


> Why ? I mean we have maybe only 5% of our customers on the us-east =
zone, what in C* require to have the same number of node on each DC ?
Because you are going to replicate your data 3 times in each DC so that =
each DC can operate with a LOCAL_QUOURM.=20

> What is better on adding nodes with no data and then rebuild them =
compared to using the auto_bootstrap ?

nodetool rebuild is designed to handle pulling data from another dc, so =
you can use it when the local DC does not contain data. i.e. you do not =
want a node in the new DC bootstrapping from other nodes in the new DC, =
they have no data.=20
 =20
> Any doc on this ? I am not aware of all the possibilities. Why is this =
the best method according to you ?
=
http://wiki.apache.org/cassandra/Operations?highlight=3D%28token%29#Token_=
selection
http://www.datastax.com/docs/1.2/initialize/token_generation

Cause it's easier to understand than interleaving the nodes and works =
with 2+ DC's.=20

> What is the point of this ?
http://wiki.apache.org/cassandra/FAQ#seed

> I didn't thought this change would be that tricky, thank you guys for =
these warnings and your help ;)
Yup, this is a lot of work.=20

Cheers
-----------------
Aaron Morton
Freelance Cassandra Consultant
New Zealand

@aaronmorton
http://www.thelastpickle.com

On 23/04/2013, at 7:26 PM, Alain RODRIGUEZ <arodrime@gmail.com> wrote:

> Hi,these advice are very welcome.
>=20
> @Dane, about the rack awareness, we use only one rack per DC, so I =
guess using EC2MultiRegionSnitch will do just fine and it doesn't need =
any configuration. Does it seem right to you. If we are someday =
interested on multi racks I will make sure to use them properly. Thank =
you for this insight anyway. You are advising me to test it, what would =
be a good way of testing it (I can use AWS EC2 instances if needed) ?
>=20
> @Aaron
>=20
> "I recommend using the same number of nodes in both DC's."
>=20
> Why ? I mean we have maybe only 5% of our customers on the us-east =
zone, what in C* require to have the same number of node on each DC ?
>=20
> "Add the nodes (I recommend 6) with auto_bootstrap: false added to the =
yaml.
> update the keyspace replication strategy to add rf:3 for the new DC.=20=

> Use nodetool rebuild on the new nodes to rebuild them from the us-west =
DC. "
>=20
> What is better on adding nodes with no data and then rebuild them =
compared to using the auto_bootstrap ?
>=20
> "I prefer to use the offset method. Take the 6 tokens from your =
us-west DC and add 100 to them for the new dc. "
>=20
> Any doc on this ? I am not aware of all the possibilities. Why is this =
the best method according to you ?
>=20
> About seeds =3D> "Yes. Have 3 from each."
>=20
> What is the point of this ?
>=20
> I didn't thought this change would be that tricky, thank you guys for =
these warnings and your help ;)
>=20
> Alain
>=20
>=20
> 2013/4/23 Dane Miller <dane@optimalsocial.com>
> On Thu, Apr 18, 2013 at 7:41 AM, Alain RODRIGUEZ <arodrime@gmail.com> =
wrote:
> > I am wondering about the process to grow from one data center to a =
few of
> > them. First thing is we use EC2Snitch for now. So I guess we have to =
switch
> > to Ec2MultiRegionSnitch.
> >
> > c/ I am using the SimpleStrategy. Is it worth it/mandatory to change =
this
> > strategy when using multiple DC ?
>=20
> I suggest you thoroughly read the datastax documentation on cassandra
> replication.  The change you are planning is big - make sure to try it
> in a test environment first.  Also, you might find you don't really
> need Cassandra's rack aware feature, and can operate using
> (Gossiping)PropertyFileSnitch.  The rack feature is listed as an
> "anti-pattern" here:
> http://www.datastax.com/docs/1.2/cluster_architecture/anti_patterns
>=20
> Here are some recent discussions on this list:
> =
http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/migrating=
-from-SimpleStrategy-to-NetworkTopologyStrategy-tp7586272.html
> =
http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/migrating=
-from-SimpleStrategy-to-NetworkTopologyStrategy-tp7481090.html
>=20
> Dane
>=20