Endless recursive call when the ZK quorum spec is invalid about asynchbase HOT 5 CLOSED

Thats the InetAddress implementation? How is that? They are different throws?
St.Ack

On Wed, May 11, 2011 at 11:15 PM, tsuna
[email protected]
wrote:

Give asynchbase a single hostname that doesn't exist for a ZooKeeper quorum spec and observe this:

2011-05-11 23:09:52,631 ERROR [main] HBaseClient: Failed to connect to ZooKeeper
java.net.UnknownHostException: tmpzookeeper
       at java.net.Inet6AddressImpl.lookupAllHostAddr(Native Method) ~[na:1.6.0_24]
       at java.net.InetAddress$1.lookupAllHostAddr(InetAddress.java:850) ~[na:1.6.0_24]
       at java.net.InetAddress.getAddressFromNameService(InetAddress.java:1201) ~[na:1.6.0_24]
       at java.net.InetAddress.getAllByName0(InetAddress.java:1154) ~[na:1.6.0_24]
       at java.net.InetAddress.getAllByName(InetAddress.java:1084) ~[na:1.6.0_24]
       at java.net.InetAddress.getAllByName(InetAddress.java:1020) ~[na:1.6.0_24]
       at org.apache.zookeeper.ClientCnxn.(ClientCnxn.java:383) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.apache.zookeeper.ClientCnxn.(ClientCnxn.java:328) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.apache.zookeeper.ZooKeeper.(ZooKeeper.java:377) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2203) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.getDeferredRoot(HBaseClient.java:2120) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.locateRegion(HBaseClient.java:1102) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.sendRpcToRegion(HBaseClient.java:1005) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.get(HBaseClient.java:700) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.ensureTableFamilyExists(HBaseClient.java:653) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.ensureTableExists(HBaseClient.java:673) [hbaseasync-1.0.jar:d1aff70]
       at net.opentsdb.tools.TSDMain.main(TSDMain.java:130) [tsdb-1.0.jar:b4a37c5]
[...]
2011-05-11 23:09:52,644 ERROR [main] HBaseClient: Failed to connect to ZooKeeper
java.net.UnknownHostException: tmpzookeeper
       at java.net.InetAddress.getAllByName0(InetAddress.java:1158) ~[na:1.6.0_24]
       at java.net.InetAddress.getAllByName(InetAddress.java:1084) ~[na:1.6.0_24]
       at java.net.InetAddress.getAllByName(InetAddress.java:1020) ~[na:1.6.0_24]
       at org.apache.zookeeper.ClientCnxn.(ClientCnxn.java:383) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.apache.zookeeper.ClientCnxn.(ClientCnxn.java:328) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.apache.zookeeper.ZooKeeper.(ZooKeeper.java:377) ~[zookeeper-3.3.1.jar:3.3.1-942149]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2203) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.connectZK(HBaseClient.java:2209) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient$ZKClient.getDeferredRoot(HBaseClient.java:2120) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.locateRegion(HBaseClient.java:1102) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.sendRpcToRegion(HBaseClient.java:1005) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.get(HBaseClient.java:700) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.ensureTableFamilyExists(HBaseClient.java:653) [hbaseasync-1.0.jar:d1aff70]
       at org.hbase.async.HBaseClient.ensureTableExists(HBaseClient.java:673) [hbaseasync-1.0.jar:d1aff70]
       at net.opentsdb.tools.TSDMain.main(TSDMain.java:130) [tsdb-1.0.jar:b4a37c5]

Notice how stack frames in connectZK keep adding up?  Bad.

Reply to this email directly or view it on GitHub:
https://github.com/stumbleupon/asynchbase/issues/5

from asynchbase.

No the problem is that ZKClient.connectZK is calling itself recursively when it catches an IOException right after creating a ZooKeeper object. The code is roughly like this:

    private void connectZK() {
      // [...]
        // Session establishment is asynchronous, so this won't block.
        zk = new ZooKeeper(quorum_spec, 5000, this);
      } catch (IOException e) {
        LOG.error("Failed to connect to ZooKeeper", e);
        // XXX don't retry recursively, create a timer with an exponential
        // backoff and schedule the reconnection attempt for later.
        connectZK();
      }

It's time to fix that XXX comment! :)

from asynchbase.

good one. you told the boys about it?

On Thu, May 12, 2011 at 8:43 AM, tsuna
[email protected]
wrote:

No the problem is that ZKClient.connectZK is calling itself recursively when it catches an IOException right after creating a ZooKeeper object.  The code is roughly like this:

   private void connectZK() {
     // [...]
       // Session establishment is asynchronous, so this won't block.
       zk = new ZooKeeper(quorum_spec, 5000, this);
     } catch (IOException e) {
       LOG.error("Failed to connect to ZooKeeper", e);
       // XXX don't retry recursively, create a timer with an exponential
       // backoff and schedule the reconnection attempt for later.
       connectZK();
     }

It's time to fix that XXX comment! :)

Reply to this email directly or view it on GitHub:
https://github.com/stumbleupon/asynchbase/issues/5#comment_1147334

from asynchbase.

... if only java had tail-call elimination. sigh

from asynchbase.

Although it's really too bad that the JVM platform (or rather the byte code) doesn't do tail-call optimizations, in this case it would simply turn the imminent stack overflow error into an endless loop.

from asynchbase.