Heka doesn't yet have any built in transport failover, but you can use
HAProxy to handle this for TCP connections.
-r
On 11/03/2014 05:04 AM, Gerhard Häring wrote:
Hello there,
I'm just trying to set up a Heka log delivery solution as an alternative
to the Logstash based one I currently have. The web nodes are connected
via TcpOutput to es.my.domain, which points to the healthy
hekad/ElasticSearch receiver hosts. I tested wether the node's hekad
will switch over to a healthy host when i stop the "receiver" hekad.
This works.
es.my.domain
A es01.my.domain (active/inactive via route53 health check)
A es02.my.domain (active/inactive via route53 health check)
A es03.my.domain (active/inactive via route53 health check)
es01.my.domain es02.my.domain es03.my.domain
+--+--------------+-+ +--+--------------+-+
+--+--------------+-+
| +--------------+ | | +--------------+ | |
+--------------+ |
| |hekad TcpInput| | | |hekad TcpInput| | | |hekad
TcpInput| |
| |TLS | | | |TLS | | | |TLS
| |
| +------+-------+ | | +------+-------+ | |
+------+-------+ |
| | | | | | | |
|
| | | | | | | |
|
| +------v-------+ | | +------v-------+ | |
+------v-------+ |
| |ElasticSearch | | | |ElasticSearch | | | |ElasticSearch
| |
| +--------------+ | | +--------------+ | |
+--------------+ |
+--+--------------+-+ +--+--------------+-+
+--+--------------+-+
My question is if you have any other suggestions on how to build a setup
that is resilient to one of the ElasticSearch/hekad hosts going down.
I know that using RabbitMQ via the amqp output is an option, but are
there any examples for a RabbitMQ connection string for more than one host?
I'd also prefer to avoid having to set up a RabbitMQ cluster if possible.
-- Gerhard
_______________________________________________
Heka mailing list
[email protected]
https://mail.mozilla.org/listinfo/heka
_______________________________________________
Heka mailing list
[email protected]
https://mail.mozilla.org/listinfo/heka
