Skip to content

Commit 2e70890

Browse files
DaveCTurnerDaveCTurner
committed
Move file-based discovery to core (#33241)
Today we support a static list of seed hosts in core Elasticsearch, and allow a dynamic list of seed hosts to be provided via a file using the `discovery-file` plugin. In fact the ability to provide a dynamic list of seed hosts is increasingly useful, so this change moves this functionality to core Elasticsearch to avoid the need for a plugin. Furthermore, in order to start up nodes in integration tests we currently assign a known port to each node before startup, which unfortunately sometimes fails if another process grabs the selected port in the meantime. By moving the `discovery-file` functionality into the core product we can use it to avoid this race. This change also moves the expected path to the file from `$ES_PATH_CONF/discovery-file/unicast_hosts.txt` to `$ES_PATH_CONF/unicast_hosts.txt`. An example of this file is not included in distributions. For BWC purposes the plugin still exists, but does nothing more than create the example file in the old location, and issue a warning when it is used. We also continue to support the old location for the file, but warn about its deprecation. Relates #29244 Closes #33030
1 parent 39244bb commit 2e70890

File tree

10 files changed

+370
-290
lines changed

10 files changed

+370
-290
lines changed

‎docs/plugins/discovery-file.asciidoc

+8-65
Original file line numberDiff line numberDiff line change
@@ -1,71 +1,14 @@
11
[[discovery-file]]
22
=== File-Based Discovery Plugin
33

4-
The file-based discovery plugin uses a list of hosts/ports in a `unicast_hosts.txt` file
5-
in the `config/discovery-file` directory for unicast discovery.
4+
The functionality provided by the `discovery-file` plugin is now available in
5+
Elasticsearch without requiring a plugin. This plugin still exists to ensure
6+
backwards compatibility, but it will be removed in a future version.
7+
8+
On installation, this plugin creates a file at
9+
`$ES_PATH_CONF/discovery-file/unicast_hosts.txt` that comprises comments that
10+
describe how to use it. It is preferable not to install this plugin and instead
11+
to create this file, and its containing directory, using standard tools.
612

713
:plugin_name: discovery-file
814
include::install_remove.asciidoc[]
9-
10-
[[discovery-file-usage]]
11-
[float]
12-
==== Using the file-based discovery plugin
13-
14-
The file-based discovery plugin provides the ability to specify the
15-
unicast hosts list through a simple `unicast_hosts.txt` file that can
16-
be dynamically updated at any time. To enable, add the following in `elasticsearch.yml`:
17-
18-
[source,yaml]
19-
----
20-
discovery.zen.hosts_provider: file
21-
----
22-
23-
This plugin simply provides a facility to supply the unicast hosts list for
24-
zen discovery through an external file that can be updated at any time by a side process.
25-
26-
For example, this gives a convenient mechanism for an Elasticsearch instance
27-
that is run in docker containers to be dynamically supplied a list of IP
28-
addresses to connect to for zen discovery when those IP addresses may not be
29-
known at node startup.
30-
31-
Note that the file-based discovery plugin is meant to augment the unicast
32-
hosts list in `elasticsearch.yml` (if specified), not replace it. Therefore,
33-
if there are valid unicast host entries in `discovery.zen.ping.unicast.hosts`,
34-
they will be used in addition to those supplied in `unicast_hosts.txt`.
35-
36-
Anytime a change is made to the `unicast_hosts.txt` file, even as Elasticsearch
37-
continues to run, the new changes will be picked up by the plugin and the
38-
new hosts list will be used for the next pinging round for master election.
39-
40-
Upon installation of the plugin, a default `unicast_hosts.txt` file will
41-
be found in the `$CONFIG_DIR/discovery-file` directory. This default file
42-
will contain some comments about what the file should contain. All comments
43-
for this file must appear on their lines starting with `#` (i.e. comments
44-
cannot start in the middle of a line).
45-
46-
[[discovery-file-format]]
47-
[float]
48-
==== unicast_hosts.txt file format
49-
50-
The format of the file is to specify one unicast host entry per line.
51-
Each unicast host entry consists of the host (host name or IP address) and
52-
an optional transport port number. If the port number is specified, is must
53-
come immediately after the host (on the same line) separated by a `:`.
54-
If the port number is not specified, a default value of 9300 is used.
55-
56-
For example, this is an example of `unicast_hosts.txt` for a cluster with
57-
four nodes that participate in unicast discovery, some of which are not
58-
running on the default port:
59-
60-
[source,txt]
61-
----------------------------------------------------------------
62-
10.10.10.5
63-
10.10.10.6:9305
64-
10.10.10.5:10005
65-
# an IPv6 address
66-
[2001:0db8:85a3:0000:0000:8a2e:0370:7334]:9301
67-
----------------------------------------------------------------
68-
69-
Host names are allowed instead of IP addresses (similar to
70-
`discovery.zen.ping.unicast.hosts`), and IPv6 addresses must be
71-
specified in brackets with the port coming after the brackets.

‎docs/reference/modules/discovery/zen.asciidoc

+160-82
Original file line numberDiff line numberDiff line change
@@ -1,100 +1,172 @@
11
[[modules-discovery-zen]]
22
=== Zen Discovery
33

4-
The zen discovery is the built in discovery module for Elasticsearch and
5-
the default. It provides unicast discovery, but can be extended to
6-
support cloud environments and other forms of discovery.
4+
Zen discovery is the built-in, default, discovery module for Elasticsearch. It
5+
provides unicast and file-based discovery, and can be extended to support cloud
6+
environments and other forms of discovery via plugins.
77

8-
The zen discovery is integrated with other modules, for example, all
9-
communication between nodes is done using the
10-
<<modules-transport,transport>> module.
8+
Zen discovery is integrated with other modules, for example, all communication
9+
between nodes is done using the <<modules-transport,transport>> module.
1110

1211
It is separated into several sub modules, which are explained below:
1312

1413
[float]
1514
[[ping]]
1615
==== Ping
1716

18-
This is the process where a node uses the discovery mechanisms to find
19-
other nodes.
17+
This is the process where a node uses the discovery mechanisms to find other
18+
nodes.
19+
20+
[float]
21+
[[discovery-seed-nodes]]
22+
==== Seed nodes
23+
24+
Zen discovery uses a list of _seed_ nodes in order to start off the discovery
25+
process. At startup, or when electing a new master, Elasticsearch tries to
26+
connect to each seed node in its list, and holds a gossip-like conversation with
27+
them to find other nodes and to build a complete picture of the cluster. By
28+
default there are two methods for configuring the list of seed nodes: _unicast_
29+
and _file-based_. It is recommended that the list of seed nodes comprises the
30+
list of master-eligible nodes in the cluster.
2031

2132
[float]
2233
[[unicast]]
2334
===== Unicast
2435

25-
Unicast discovery requires a list of hosts to use that will act as gossip
26-
routers. These hosts can be specified as hostnames or IP addresses; hosts
27-
specified as hostnames are resolved to IP addresses during each round of
28-
pinging. Note that if you are in an environment where DNS resolutions vary with
29-
time, you might need to adjust your <<networkaddress-cache-ttl,JVM security
30-
settings>>.
36+
Unicast discovery configures a static list of hosts for use as seed nodes.
37+
These hosts can be specified as hostnames or IP addresses; hosts specified as
38+
hostnames are resolved to IP addresses during each round of pinging. Note that
39+
if you are in an environment where DNS resolutions vary with time, you might
40+
need to adjust your <<networkaddress-cache-ttl,JVM security settings>>.
3141

32-
It is recommended that the unicast hosts list be maintained as the list of
33-
master-eligible nodes in the cluster.
42+
The list of hosts is set using the `discovery.zen.ping.unicast.hosts` static
43+
setting. This is either an array of hosts or a comma-delimited string. Each
44+
value should be in the form of `host:port` or `host` (where `port` defaults to
45+
the setting `transport.profiles.default.port` falling back to
46+
`transport.tcp.port` if not set). Note that IPv6 hosts must be bracketed. The
47+
default for this setting is `127.0.0.1, [::1]`
3448

35-
Unicast discovery provides the following settings with the `discovery.zen.ping.unicast` prefix:
49+
Additionally, the `discovery.zen.ping.unicast.resolve_timeout` configures the
50+
amount of time to wait for DNS lookups on each round of pinging. This is
51+
specified as a <<time-units, time unit>> and defaults to 5s.
3652

37-
[cols="<,<",options="header",]
38-
|=======================================================================
39-
|Setting |Description
40-
|`hosts` |Either an array setting or a comma delimited setting. Each
41-
value should be in the form of `host:port` or `host` (where `port` defaults to the setting `transport.profiles.default.port`
42-
falling back to `transport.tcp.port` if not set). Note that IPv6 hosts must be bracketed. Defaults to `127.0.0.1, [::1]`
43-
|`hosts.resolve_timeout` |The amount of time to wait for DNS lookups on each round of pinging. Specified as
44-
<<time-units, time units>>. Defaults to 5s.
45-
|=======================================================================
53+
Unicast discovery uses the <<modules-transport,transport>> module to perform the
54+
discovery.
4655

47-
The unicast discovery uses the <<modules-transport,transport>> module to perform the discovery.
56+
[float]
57+
[[file-based-hosts-provider]]
58+
===== File-based
59+
60+
In addition to hosts provided by the static `discovery.zen.ping.unicast.hosts`
61+
setting, it is possible to provide a list of hosts via an external file.
62+
Elasticsearch reloads this file when it changes, so that the list of seed nodes
63+
can change dynamically without needing to restart each node. For example, this
64+
gives a convenient mechanism for an Elasticsearch instance that is run in a
65+
Docker container to be dynamically supplied with a list of IP addresses to
66+
connect to for Zen discovery when those IP addresses may not be known at node
67+
startup.
68+
69+
To enable file-based discovery, configure the `file` hosts provider as follows:
70+
71+
```
72+
discovery.zen.hosts_provider: file
73+
```
74+
75+
Then create a file at `$ES_PATH_CONF/unicast_hosts.txt` in
76+
<<discovery-file-format,the format described below>>. Any time a change is made
77+
to the `unicast_hosts.txt` file the new changes will be picked up by
78+
Elasticsearch and the new hosts list will be used.
79+
80+
Note that the file-based discovery plugin augments the unicast hosts list in
81+
`elasticsearch.yml`: if there are valid unicast host entries in
82+
`discovery.zen.ping.unicast.hosts` then they will be used in addition to those
83+
supplied in `unicast_hosts.txt`.
84+
85+
The `discovery.zen.ping.unicast.resolve_timeout` setting also applies to DNS
86+
lookups for nodes specified by address via file-based discovery. This is
87+
specified as a <<time-units, time unit>> and defaults to 5s.
88+
89+
[[discovery-file-format]]
90+
[float]
91+
====== unicast_hosts.txt file format
92+
93+
The format of the file is to specify one node entry per line. Each node entry
94+
consists of the host (host name or IP address) and an optional transport port
95+
number. If the port number is specified, is must come immediately after the
96+
host (on the same line) separated by a `:`. If the port number is not
97+
specified, a default value of 9300 is used.
98+
99+
For example, this is an example of `unicast_hosts.txt` for a cluster with four
100+
nodes that participate in unicast discovery, some of which are not running on
101+
the default port:
102+
103+
[source,txt]
104+
----------------------------------------------------------------
105+
10.10.10.5
106+
10.10.10.6:9305
107+
10.10.10.5:10005
108+
# an IPv6 address
109+
[2001:0db8:85a3:0000:0000:8a2e:0370:7334]:9301
110+
----------------------------------------------------------------
111+
112+
Host names are allowed instead of IP addresses (similar to
113+
`discovery.zen.ping.unicast.hosts`), and IPv6 addresses must be specified in
114+
brackets with the port coming after the brackets.
115+
116+
It is also possible to add comments to this file. All comments must appear on
117+
their lines starting with `#` (i.e. comments cannot start in the middle of a
118+
line).
48119

49120
[float]
50121
[[master-election]]
51122
==== Master Election
52123

53-
As part of the ping process a master of the cluster is either
54-
elected or joined to. This is done automatically. The
55-
`discovery.zen.ping_timeout` (which defaults to `3s`) determines how long the node
56-
will wait before deciding on starting an election or joining an existing cluster.
57-
Three pings will be sent over this timeout interval. In case where no decision can be
58-
reached after the timeout, the pinging process restarts.
59-
In slow or congested networks, three seconds might not be enough for a node to become
60-
aware of the other nodes in its environment before making an election decision.
61-
Increasing the timeout should be done with care in that case, as it will slow down the
62-
election process.
63-
Once a node decides to join an existing formed cluster, it
64-
will send a join request to the master (`discovery.zen.join_timeout`)
65-
with a timeout defaulting at 20 times the ping timeout.
66-
67-
When the master node stops or has encountered a problem, the cluster nodes
68-
start pinging again and will elect a new master. This pinging round also
69-
serves as a protection against (partial) network failures where a node may unjustly
70-
think that the master has failed. In this case the node will simply hear from
71-
other nodes about the currently active master.
72-
73-
If `discovery.zen.master_election.ignore_non_master_pings` is `true`, pings from nodes that are not master
74-
eligible (nodes where `node.master` is `false`) are ignored during master election; the default value is
124+
As part of the ping process a master of the cluster is either elected or joined
125+
to. This is done automatically. The `discovery.zen.ping_timeout` (which defaults
126+
to `3s`) determines how long the node will wait before deciding on starting an
127+
election or joining an existing cluster. Three pings will be sent over this
128+
timeout interval. In case where no decision can be reached after the timeout,
129+
the pinging process restarts. In slow or congested networks, three seconds
130+
might not be enough for a node to become aware of the other nodes in its
131+
environment before making an election decision. Increasing the timeout should
132+
be done with care in that case, as it will slow down the election process. Once
133+
a node decides to join an existing formed cluster, it will send a join request
134+
to the master (`discovery.zen.join_timeout`) with a timeout defaulting at 20
135+
times the ping timeout.
136+
137+
When the master node stops or has encountered a problem, the cluster nodes start
138+
pinging again and will elect a new master. This pinging round also serves as a
139+
protection against (partial) network failures where a node may unjustly think
140+
that the master has failed. In this case the node will simply hear from other
141+
nodes about the currently active master.
142+
143+
If `discovery.zen.master_election.ignore_non_master_pings` is `true`, pings from
144+
nodes that are not master eligible (nodes where `node.master` is `false`) are
145+
ignored during master election; the default value is `false`.
146+
147+
Nodes can be excluded from becoming a master by setting `node.master` to
75148
`false`.
76149

77-
Nodes can be excluded from becoming a master by setting `node.master` to `false`.
78-
79-
The `discovery.zen.minimum_master_nodes` sets the minimum
80-
number of master eligible nodes that need to join a newly elected master in order for an election to
81-
complete and for the elected node to accept its mastership. The same setting controls the minimum number of
82-
active master eligible nodes that should be a part of any active cluster. If this requirement is not met the
83-
active master node will step down and a new master election will begin.
150+
The `discovery.zen.minimum_master_nodes` sets the minimum number of master
151+
eligible nodes that need to join a newly elected master in order for an election
152+
to complete and for the elected node to accept its mastership. The same setting
153+
controls the minimum number of active master eligible nodes that should be a
154+
part of any active cluster. If this requirement is not met the active master
155+
node will step down and a new master election will begin.
84156

85157
This setting must be set to a <<minimum_master_nodes,quorum>> of your master
86158
eligible nodes. It is recommended to avoid having only two master eligible
87-
nodes, since a quorum of two is two. Therefore, a loss of either master
88-
eligible node will result in an inoperable cluster.
159+
nodes, since a quorum of two is two. Therefore, a loss of either master eligible
160+
node will result in an inoperable cluster.
89161

90162
[float]
91163
[[fault-detection]]
92164
==== Fault Detection
93165

94-
There are two fault detection processes running. The first is by the
95-
master, to ping all the other nodes in the cluster and verify that they
96-
are alive. And on the other end, each node pings to master to verify if
97-
its still alive or an election process needs to be initiated.
166+
There are two fault detection processes running. The first is by the master, to
167+
ping all the other nodes in the cluster and verify that they are alive. And on
168+
the other end, each node pings to master to verify if its still alive or an
169+
election process needs to be initiated.
98170

99171
The following settings control the fault detection process using the
100172
`discovery.zen.fd` prefix:
@@ -116,19 +188,21 @@ considered failed. Defaults to `3`.
116188

117189
The master node is the only node in a cluster that can make changes to the
118190
cluster state. The master node processes one cluster state update at a time,
119-
applies the required changes and publishes the updated cluster state to all
120-
the other nodes in the cluster. Each node receives the publish message, acknowledges
121-
it, but does *not* yet apply it. If the master does not receive acknowledgement from
122-
at least `discovery.zen.minimum_master_nodes` nodes within a certain time (controlled by
123-
the `discovery.zen.commit_timeout` setting and defaults to 30 seconds) the cluster state
124-
change is rejected.
125-
126-
Once enough nodes have responded, the cluster state is committed and a message will
127-
be sent to all the nodes. The nodes then proceed to apply the new cluster state to their
128-
internal state. The master node waits for all nodes to respond, up to a timeout, before
129-
going ahead processing the next updates in the queue. The `discovery.zen.publish_timeout` is
130-
set by default to 30 seconds and is measured from the moment the publishing started. Both
131-
timeout settings can be changed dynamically through the <<cluster-update-settings,cluster update settings api>>
191+
applies the required changes and publishes the updated cluster state to all the
192+
other nodes in the cluster. Each node receives the publish message, acknowledges
193+
it, but does *not* yet apply it. If the master does not receive acknowledgement
194+
from at least `discovery.zen.minimum_master_nodes` nodes within a certain time
195+
(controlled by the `discovery.zen.commit_timeout` setting and defaults to 30
196+
seconds) the cluster state change is rejected.
197+
198+
Once enough nodes have responded, the cluster state is committed and a message
199+
will be sent to all the nodes. The nodes then proceed to apply the new cluster
200+
state to their internal state. The master node waits for all nodes to respond,
201+
up to a timeout, before going ahead processing the next updates in the queue.
202+
The `discovery.zen.publish_timeout` is set by default to 30 seconds and is
203+
measured from the moment the publishing started. Both timeout settings can be
204+
changed dynamically through the <<cluster-update-settings,cluster update
205+
settings api>>
132206

133207
[float]
134208
[[no-master-block]]
@@ -143,10 +217,14 @@ rejected when there is no active master.
143217
The `discovery.zen.no_master_block` setting has two valid options:
144218

145219
[horizontal]
146-
`all`:: All operations on the node--i.e. both read & writes--will be rejected. This also applies for api cluster state
147-
read or write operations, like the get index settings, put mapping and cluster state api.
148-
`write`:: (default) Write operations will be rejected. Read operations will succeed, based on the last known cluster configuration.
149-
This may result in partial reads of stale data as this node may be isolated from the rest of the cluster.
150-
151-
The `discovery.zen.no_master_block` setting doesn't apply to nodes-based apis (for example cluster stats, node info and
152-
node stats apis). Requests to these apis will not be blocked and can run on any available node.
220+
`all`:: All operations on the node--i.e. both read & writes--will be rejected.
221+
This also applies for api cluster state read or write operations, like the get
222+
index settings, put mapping and cluster state api.
223+
`write`:: (default) Write operations will be rejected. Read operations will
224+
succeed, based on the last known cluster configuration. This may result in
225+
partial reads of stale data as this node may be isolated from the rest of the
226+
cluster.
227+
228+
The `discovery.zen.no_master_block` setting doesn't apply to nodes-based apis
229+
(for example cluster stats, node info and node stats apis). Requests to these
230+
apis will not be blocked and can run on any available node.

0 commit comments

Comments
 (0)