[
https://issues.apache.org/jira/browse/IGNITE-13206?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17225362#comment-17225362
]
Vladimir Steshin edited comment on IGNITE-13206 at 11/3/20, 6:39 PM:
---------------------------------------------------------------------
[~dmagda], wrong. It might be confusion with addresses to detect other nodes to
connect the ring in the IP Finder. The ticket is exactly about Discovery SPI.
The problem remains as described: several IP addresses node uses (not search)
can prolog detection of node failure. We also should represent this issue in
the documentation. Users often do not set IP address of node and allow
automatically pick up all available addresses. This is lack of configuration I
think.
All node addresses of any node are saved and sent with node info. If node
doesn't respond, other node will try to connect to other provided address of
not responding node.
Better doc is about localAddress.
was (Author: vladsz83):
[~dmagda], wrong. It might be confusion with addresses to detect other nodes to
connect the ring in the IP Finder. The ticket is exactly about Discovery SPI.
The problem remains as described: several IP addresses node uses (not search)
can prolog detection of node failure. Similar ticket is IGNITE-13205. But it
warns in Javadoc and logs. We also should represent this issue in the
documentation. Users often do not set IP address of node and allow
automatically pick up all available addresses. This is lack of configuration I
think.
All node addresses of any node are saved and sent with node info. If node
doesn't respond, other node will try to connect to other provided address of
not responding node.
Better doc is about localAddress.
> Represent in the documenttion affection of several node addresses on failure
> detection.
> ---------------------------------------------------------------------------------------
>
> Key: IGNITE-13206
> URL: https://issues.apache.org/jira/browse/IGNITE-13206
> Project: Ignite
> Issue Type: Improvement
> Components: documentation
> Reporter: Vladimir Steshin
> Assignee: Denis A. Magda
> Priority: Major
> Labels: iep-45
> Fix For: 2.10
>
>
> We should emphasize that TcpDiscoverySpi prolongs detection of node failure
> if several IP addresses are set. Actual failure detection delay is:
> _failureDetectionTimeout * addressesNumber_.
> "You should assing multiple addresses to a node only if they represent some
> real physical connections which can give more reliability. Several addresses
> prolong failure detection of current node. The timeouts and settings on
> network operations (_failureDetectionTimeout(), sockTimeout, ackTimeout,
> maxAckTimeout, reconCnt_) work per connection/address. The exception is
> _connRecoveryTimeout_.
> Example: if you have 3 ip addresses configured for a node, Tcp Discovery
> takes up to '_failureDetectionTimeout * 3 + connRecoveryTimeout' to detect
> failure of this node_".
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
