Hello,
We use one ES cluster with 4 nodes spread across 2 data centers (2
nodes/DC). Each DC is configured as a zone
(via cluster.routing.allocation.awareness.attributes).
I would like to use snapshot to backup indexes using type:fs. Per http://www.elastic.co/guide/en/elasticsearch/reference/current/modules-snapshots.html,
the requirement is to have a location that is accessible on all data and
master nodes. Unfortunately, I don't have a location that is accessible
from both DCs (cross-mounts across DCs is not supported).
Is there a way to backup indexes in a given DC/zone, especially given that
all the primary and replica shards are available in a DC anyway?
Hello,
We use one ES cluster with 4 nodes spread across 2 data centers (2
nodes/DC). Each DC is configured as a zone
(via cluster.routing.allocation.awareness.attributes).
I would like to use snapshot to backup indexes using type:fs. Per Snapshot module | Elasticsearch Guide [8.11] | Elastic,
the requirement is to have a location that is accessible on all data and
master nodes. Unfortunately, I don't have a location that is accessible
from both DCs (cross-mounts across DCs is not supported).
Is there a way to backup indexes in a given DC/zone, especially given that
all the primary and replica shards are available in a DC anyway?
Thanks. I tried using a local mount point and I have backup created on all
4 nodes because the primary shards are spread across all nodes. I am
guessing that I need all 4 nodes online whenever I need to, for e.g.,
restore a snapshot.
The latency between the DCs is <2ms, so for all practical reasons, we often
don't distinguish the DCs. Other than the latency, what other stuff that I
should be concerned about w.r.to ES cluster spread across DCs?
On Friday, March 13, 2015 at 12:04:23 AM UTC-4, Mark Walkom wrote:
Just use a local mount point, as long as the path is the same it doesn't
matter.
Also, we do not recommend cross DC clusters, there is a lot of potential
issues you can run into.
On 12 March 2015 at 17:31, Foobar Geez <fooba...@gmail.com <javascript:>>
wrote:
Hello,
We use one ES cluster with 4 nodes spread across 2 data centers (2
nodes/DC). Each DC is configured as a zone
(via cluster.routing.allocation.awareness.attributes).
I would like to use snapshot to backup indexes using type:fs. Per Snapshot module | Elasticsearch Guide [8.11] | Elastic,
the requirement is to have a location that is accessible on all data and
master nodes. Unfortunately, I don't have a location that is accessible
from both DCs (cross-mounts across DCs is not supported).
Is there a way to backup indexes in a given DC/zone, especially given
that all the primary and replica shards are available in a DC anyway?
The second one is that if your link drops then chances are you will end up
with a split brain, unless you have set min masters correctly, or a
completely inaccessible cluster.
Thanks. I tried using a local mount point and I have backup created on
all 4 nodes because the primary shards are spread across all nodes. I am
guessing that I need all 4 nodes online whenever I need to, for e.g.,
restore a snapshot.
The latency between the DCs is <2ms, so for all practical reasons, we
often don't distinguish the DCs. Other than the latency, what other stuff
that I should be concerned about w.r.to ES cluster spread across DCs?
On Friday, March 13, 2015 at 12:04:23 AM UTC-4, Mark Walkom wrote:
Just use a local mount point, as long as the path is the same it doesn't
matter.
Also, we do not recommend cross DC clusters, there is a lot of potential
issues you can run into.
Hello,
We use one ES cluster with 4 nodes spread across 2 data centers (2
nodes/DC). Each DC is configured as a zone (via cluster.routing.
allocation.awareness.attributes).
I would like to use snapshot to backup indexes using type:fs. Per Elasticsearch Guide | Elastic
current/modules-snapshots.html, the requirement is to have a location
that is accessible on all data and master nodes. Unfortunately, I don't
have a location that is accessible from both DCs (cross-mounts across DCs
is not supported).
Is there a way to backup indexes in a given DC/zone, especially given
that all the primary and replica shards are available in a DC anyway?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.