Seeking opinions on cluster platforms

Let me pose a question by suggesting two extremes for hardware to create
and maintain a growing ElasticSearch cluster datacenter (not in the cloud).

One extreme places redundancy at the server hardware level, by which I mean:
dual power supplies, RAID hard drives

Another extreme places redundancy in a multitude of backup servers:
commodity servers, single power supply, no RAID on the disks, low cost,
with a cluster monitor that can advise of a failed master or backup, and
can rebuild the replacement

I would love to learn how others see or implement within the boundaries of
those extremes, with the understanding that the two poles are just
suggestions, there may be other ways to slice this space.

Many thanks in advance
Jack
ps: documents I read based on a broad query:


http://www.elasticsearch.org/case-study/maptimize/
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-gateway.html

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Personally, I'd go with the latter and then let the software handle all the
redundancy. You can get super cheap 1RU pizza boxes from Quanta or the like
and save yourself a bundle in that area and then leverage automation and
configuration using The Foreman and Puppet.

Tie a bit more smarts into it and you would have an awesome elastic compute
platform. Or just use something like OpenStack, though it might be a bit
heavy.

Regards,
Mark Walkom

Infrastructure Engineer
Campaign Monitor
email: markw@campaignmonitor.com
web: www.campaignmonitor.com

On 13 September 2014 03:16, Jack Park jackpark@topicquests.org wrote:

Let me pose a question by suggesting two extremes for hardware to create
and maintain a growing ElasticSearch cluster datacenter (not in the cloud).

One extreme places redundancy at the server hardware level, by which I
mean:
dual power supplies, RAID hard drives

Another extreme places redundancy in a multitude of backup servers:
commodity servers, single power supply, no RAID on the disks, low cost,
with a cluster monitor that can advise of a failed master or backup, and
can rebuild the replacement

I would love to learn how others see or implement within the boundaries of
those extremes, with the understanding that the two poles are just
suggestions, there may be other ways to slice this space.

Many thanks in advance
Jack
ps: documents I read based on a broad query:
https://github.com/aphyr/partitions-post
http://www.elasticsearch.org/case-study/maptimize/

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-gateway.html

http://www.slideshare.net/clintongormley/scaling-realtime-search-and-analytics-with-elasticsearch

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com?utm_medium=email&utm_source=footer
.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAEM624bho%2BcJStsAAMvx5ZMApNEqCSz3a4oEofrU7VfEeuVX%2Bg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Not sure what is "extreme". The design of ES may be a surprise for those
who are not familiar with distributed system architecture.

ES can handle faults in software. I pile up cheap 1U rack servers with 32
cores, 64G RAM, ~1TB RAID 0. All nodes are equally provisioned.

If a server fails, mostly spindle drives or fans, it is decommissioned and
repaired.

No need to monitor for master failure or making backups. Master is switched
over automatically by ES, and replica level 1 (or higher) is a must.

Jörg

On Sat, Sep 13, 2014 at 12:37 AM, Mark Walkom markw@campaignmonitor.com
wrote:

Personally, I'd go with the latter and then let the software handle all
the redundancy. You can get super cheap 1RU pizza boxes from Quanta or the
like and save yourself a bundle in that area and then leverage automation
and configuration using The Foreman and Puppet.

Tie a bit more smarts into it and you would have an awesome elastic
compute platform. Or just use something like OpenStack, though it might be
a bit heavy.

Regards,
Mark Walkom

Infrastructure Engineer
Campaign Monitor
email: markw@campaignmonitor.com
web: www.campaignmonitor.com

On 13 September 2014 03:16, Jack Park jackpark@topicquests.org wrote:

Let me pose a question by suggesting two extremes for hardware to create
and maintain a growing ElasticSearch cluster datacenter (not in the cloud).

One extreme places redundancy at the server hardware level, by which I
mean:
dual power supplies, RAID hard drives

Another extreme places redundancy in a multitude of backup servers:
commodity servers, single power supply, no RAID on the disks, low cost,
with a cluster monitor that can advise of a failed master or backup, and
can rebuild the replacement

I would love to learn how others see or implement within the boundaries
of those extremes, with the understanding that the two poles are just
suggestions, there may be other ways to slice this space.

Many thanks in advance
Jack
ps: documents I read based on a broad query:
https://github.com/aphyr/partitions-post
http://www.elasticsearch.org/case-study/maptimize/

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-gateway.html

http://www.slideshare.net/clintongormley/scaling-realtime-search-and-analytics-with-elasticsearch

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com?utm_medium=email&utm_source=footer
.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAEM624bho%2BcJStsAAMvx5ZMApNEqCSz3a4oEofrU7VfEeuVX%2Bg%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAEM624bho%2BcJStsAAMvx5ZMApNEqCSz3a4oEofrU7VfEeuVX%2Bg%40mail.gmail.com?utm_medium=email&utm_source=footer
.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAKdsXoFuXTUVQrhKpueJWKPT2ws4CtXm6mB-4rGLB-0DFkC37w%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Mark and Jörg

Many thanks for these comments. In a large sense, they confirm my
intuitions.

Cheers,
Jack

On Fri, Sep 12, 2014 at 4:00 PM, joergprante@gmail.com <
joergprante@gmail.com> wrote:

Not sure what is "extreme". The design of ES may be a surprise for those
who are not familiar with distributed system architecture.

ES can handle faults in software. I pile up cheap 1U rack servers with 32
cores, 64G RAM, ~1TB RAID 0. All nodes are equally provisioned.

If a server fails, mostly spindle drives or fans, it is decommissioned and
repaired.

No need to monitor for master failure or making backups. Master is
switched over automatically by ES, and replica level 1 (or higher) is a
must.

Jörg

On Sat, Sep 13, 2014 at 12:37 AM, Mark Walkom markw@campaignmonitor.com
wrote:

Personally, I'd go with the latter and then let the software handle all
the redundancy. You can get super cheap 1RU pizza boxes from Quanta or the
like and save yourself a bundle in that area and then leverage automation
and configuration using The Foreman and Puppet.

Tie a bit more smarts into it and you would have an awesome elastic
compute platform. Or just use something like OpenStack, though it might be
a bit heavy.

Regards,
Mark Walkom

Infrastructure Engineer
Campaign Monitor
email: markw@campaignmonitor.com
web: www.campaignmonitor.com

On 13 September 2014 03:16, Jack Park jackpark@topicquests.org wrote:

Let me pose a question by suggesting two extremes for hardware to create
and maintain a growing ElasticSearch cluster datacenter (not in the cloud).

One extreme places redundancy at the server hardware level, by which I
mean:
dual power supplies, RAID hard drives

Another extreme places redundancy in a multitude of backup servers:
commodity servers, single power supply, no RAID on the disks, low cost,
with a cluster monitor that can advise of a failed master or backup, and
can rebuild the replacement

I would love to learn how others see or implement within the boundaries
of those extremes, with the understanding that the two poles are just
suggestions, there may be other ways to slice this space.

Many thanks in advance
Jack
ps: documents I read based on a broad query:
https://github.com/aphyr/partitions-post
http://www.elasticsearch.org/case-study/maptimize/

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-gateway.html

http://www.slideshare.net/clintongormley/scaling-realtime-search-and-analytics-with-elasticsearch

--
You received this message because you are subscribed to the Google
Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAH6s0fyHDhB9N8rOCwuf%2B3GR1E8xQ4aqSoQD8cYKZwo72bHw7A%40mail.gmail.com?utm_medium=email&utm_source=footer
.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAEM624bho%2BcJStsAAMvx5ZMApNEqCSz3a4oEofrU7VfEeuVX%2Bg%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAEM624bho%2BcJStsAAMvx5ZMApNEqCSz3a4oEofrU7VfEeuVX%2Bg%40mail.gmail.com?utm_medium=email&utm_source=footer
.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/CAKdsXoFuXTUVQrhKpueJWKPT2ws4CtXm6mB-4rGLB-0DFkC37w%40mail.gmail.com
https://groups.google.com/d/msgid/elasticsearch/CAKdsXoFuXTUVQrhKpueJWKPT2ws4CtXm6mB-4rGLB-0DFkC37w%40mail.gmail.com?utm_medium=email&utm_source=footer
.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAH6s0fxkMS9cm5eY7c2j2SkFFwUdvEruWCNX2U_8Zhhr%3DQ3T3A%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.