[
https://issues.apache.org/jira/browse/CASSANDRA-7363?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14025878#comment-14025878
]
Jackson Chung commented on CASSANDRA-7363:
------------------------------------------
bq. This isn't possible, it needs to know the IP. Being able to specify a
hostname is a convenience.
maybe this makes more sense if there is completely no known host at all ? iow ,
only throw the ConfigurationException if the "reloadedMap" is empty in the end?
In aws, since we are not reserving IP, it is not possible to know the ips ahead
of time. We could use a workaround as starting the instances with those
hostname first without starting C*, but, preferably not.
Another flaw about the logic of needs to know the IP is the current check
relies on
{code}
host = InetAddress.getByName(hostString);
{code}
This implies "If a literal IP address is supplied, only the validity of the
address format is checked.", ie: not necessary the existence of that IP. So it
doesn't seem to make sense that a non-existing IP will succeed; but a
non-existing hostname will fail
> PropertyFileSnitch should allow name address that does not yet exist
> --------------------------------------------------------------------
>
> Key: CASSANDRA-7363
> URL: https://issues.apache.org/jira/browse/CASSANDRA-7363
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jackson Chung
>
> when starting a new node with PropertyFileSnitch with
> cassandra-topology.properties contains an unknown host, it fails with:
> {noformat}
> ERROR [main] 2014-06-06 17:48:38,233 DatabaseDescriptor.java (line 116) Fatal
> configuration error
> org.apache.cassandra.exceptions.ConfigurationException: Error instantiating
> snitch class 'org.apache.cassandra.locator.PropertyFileSnitch'.
> at
> org.apache.cassandra.utils.FBUtilities.construct(FBUtilities.java:503)
> at
> org.apache.cassandra.config.DatabaseDescriptor.createEndpointSnitch(DatabaseDescriptor.java:506)
> at
> org.apache.cassandra.config.DatabaseDescriptor.applyConfig(DatabaseDescriptor.java:341)
> at
> org.apache.cassandra.config.DatabaseDescriptor.<clinit>(DatabaseDescriptor.java:111)
> at
> org.apache.cassandra.service.CassandraDaemon.setup(CassandraDaemon.java:155)
> at
> org.apache.cassandra.service.CassandraDaemon.activate(CassandraDaemon.java:480)
> at
> org.apache.cassandra.service.CassandraDaemon.main(CassandraDaemon.java:569)
> Caused by: org.apache.cassandra.exceptions.ConfigurationException: Unknown
> host cassandra11-staging.amz.relateiq.com
> at
> org.apache.cassandra.locator.PropertyFileSnitch.reloadConfiguration(PropertyFileSnitch.java:174)
> at
> org.apache.cassandra.locator.PropertyFileSnitch.<init>(PropertyFileSnitch.java:60)
> at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native
> Method)
> at
> sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
> at
> sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
> at java.lang.reflect.Constructor.newInstance(Constructor.java:408)
> at java.lang.Class.newInstance(Class.java:433)
> at
> org.apache.cassandra.utils.FBUtilities.construct(FBUtilities.java:488)
> ... 6 more
> Caused by: java.net.UnknownHostException:
> cassandra11-staging.amz.relateiq.com: unknown error
> at java.net.Inet4AddressImpl.lookupAllHostAddr(Native Method)
> at java.net.InetAddress$2.lookupAllHostAddr(InetAddress.java:907)
> at
> java.net.InetAddress.getAddressesFromNameService(InetAddress.java:1302)
> at java.net.InetAddress.getAllByName0(InetAddress.java:1255)
> at java.net.InetAddress.getAllByName(InetAddress.java:1171)
> at java.net.InetAddress.getAllByName(InetAddress.java:1105)
> at java.net.InetAddress.getByName(InetAddress.java:1055)
> at
> org.apache.cassandra.locator.PropertyFileSnitch.reloadConfiguration(PropertyFileSnitch.java:170)
> ... 13 more
> {noformat}
> The real impact here is we are trying to launch a number of new nodes (via
> chef) with pre-configured hostname (and among other variables). The
> additional hostname (but not yet alive) made no impact to the existing nodes,
> which is good (looks like we only catch the ConfigurationException in the
> watcher thread, but not on the initial start); but it causes new node fail to
> start. (Pretty sure if we restart an existing one, it will fail too).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
