zookeeper-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From eribeiro <...@git.apache.org>
Subject [GitHub] zookeeper pull request #173: ZOOKEEPER-2691: recreateSocketAddresses may rec...
Date Fri, 17 Feb 2017 21:05:02 GMT
Github user eribeiro commented on a diff in the pull request:

    https://github.com/apache/zookeeper/pull/173#discussion_r101845264
  
    --- Diff: src/java/main/org/apache/zookeeper/server/quorum/QuorumPeer.java ---
    @@ -181,6 +197,33 @@ public void recreateSocketAddresses() {
                 }
             }
     
    +        /**
    +         * Resolve the hostname to IP addresses, and find one reachable address.
    +         *
    +         * @param hostname the name of the host
    +         * @param timeout the time, in millseconds, before {@link InetAddress#isReachable}
    +         *                aborts
    +         * @return a reachable IP address. If no such IP address can be found,
    +         *         just return the first IP address of the hostname.
    +         *
    +         * @exception UnknownHostException
    +         */
    +        public InetAddress getReachableAddress(String hostname, int timeout) 
    +                throws UnknownHostException {
    +            InetAddress[] addresses = InetAddress.getAllByName(hostname);
    +            for (InetAddress a : addresses) {
    +                try {
    +                    if (a.isReachable(timeout)) {
    --- End diff --
    
    My main problem with this PR is that call to `isReachable(timeout)` for two reasons: 
    
    1) the most important one: `isReachable(timeout)` seems unreliable so there are plenty
cases where it returns false even tough the node is reachable or vice-versa! https://bugs.openjdk.java.net/browse/JDK-8159410
(google "InetAddress.isReachable not working" or "InetAddress.isReachable unreliable" to see
further cases).
    
    2) This timeout can add an arbitrary delay until a reachable node can be tested.
    
    IDK what a good compromise would be for both points above (leaving as it is today could
work, so no problem, even tough I am a bit concerned), but maybe we could use a solution similar
to ZOOKEEPER-2184 and return the next address in the array (using `next = ++next % addresses.length`
to prevent out of bound exceptions).
    



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message