Zookeeper connection error

2020-05-17 01:41发布

问题:

We have a standalone zookeeper setup on a dev machine. It works fine for every other dev machine except this one testdev machine.

We get this error over and over again when trying to connect to zookeeper through testdev:

2012-11-09 14:06:53,909 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@947] - Socket connection established to zk01.dev.bunchball.net/192.168.8.58:2181, initiating session
2012-11-09 14:06:53,911 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@1183] - Unable to read additional data from server sessionid 0x0, likely server has closed socket, closing socket connection and attempting reconnect
2012-11-09 14:06:55,366 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@1058] - Opening socket connection to server zk01.dev.bunchball.net/192.168.8.58:2181
2012-11-09 14:06:55,368 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@947] - Socket connection established to zk01.dev.bunchball.net/192.168.8.58:2181, initiating session
2012-11-09 14:06:55,368 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@1183] - Unable to read additional data from server sessionid 0x0, likely server has closed socket, closing socket connection and attempting reconnect
2012-11-09 14:06:57,271 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@1058] - Opening socket connection to server zk01.dev.bunchball.net/192.168.8.58:2181
2012-11-09 14:06:57,274 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@947] - Socket connection established to zk01.dev.bunchball.net/192.168.8.58:2181, initiating session
2012-11-09 14:06:57,275 - INFO  [main-SendThread(zk01.dev.bunchball.net:2181):ClientCnxn$SendThread@1183] - Unable to read additional data from server sessionid 0x0, likely server has closed socket, closing socket connection and attempting reconnect

We tried restarting the test dev machine, and also restart zookeeper host but nothing worked. We are totally confused why it works perfectly fine with other machines except this one. What might be the cause of this?

回答1:

I just have the same situation as you and I have just fixed this problem.

It is the reason that you have configured even number of zookeepers which directly result in this problem,try to change your number of zookeeper node to a odd one.

for example the original status of my zookeeper cluster is consisted of 4 nodes,then just remove one of them which result in the number of node to be 3 well ,now its ok to startup zookeeper cluster

below is the output of successfully connect to zookeeper server

2013-04-22 22:07:05,654 [myid:] - INFO  [main:ZooKeeper@438] - Initiating client connection, connectString=localhost:2181 sessionTimeout=30000 watcher=org.apache.zookeeper.ZooKeeperMain$MyWatcher@1321ed6
Welcome to ZooKeeper!
2013-04-22 22:07:05,704 [myid:] - INFO  [main-SendThread(localhost:2181):ClientCnxn$SendThread@966] - Opening socket connection to server localhost/127.0.0.1:2181. Will not attempt to authenticate using SASL (unknown error)
JLine support is enabled
2013-04-22 22:07:05,727 [myid:] - INFO  [main-SendThread(localhost:2181):ClientCnxn$SendThread@849] - Socket connection established to localhost/127.0.0.1:2181, initiating session
[zk: localhost:2181(CONNECTING) 0] 2013-04-22 22:07:05,846 [myid:] - INFO  [main-SendThread(localhost:2181):ClientCnxn$SendThread@1207] - Session establishment complete on server localhost/127.0.0.1:2181, sessionid = 0x13e3211c06e0000, negotiated timeout = 30000


回答2:

I faced the same issue and found it was due to zookeeper cluster nodes needs ports opened to communicate with each other.

server.1=xx.xx.xx.xx:2888:3888

server.2=xx.xx.xx.xx:2888:3888

server.3=xx.xx.xx.xx:2888:3888

once i allowed these ports through aws security group and restarted. All worked fine for me



回答3:

I have just solved the problem. I am using centos 7. And the trouble-maker is firewall.Using "systemctl stop firewalld" to shut it all down in each server can simply solve the problem.Or you can use command like

firewall-cmd --zone=public --add-port=2181/udp --add-port=2181/tcp --permanent" to configure all three ports ,include 2181,2888,3888 in each server.And then "firewall-cmd --reload

Finally use

zkServer.sh restart

to restart your servers and problem solved.



回答4:

In my case, I config zoo.cfg like this:

server.1=host-1:2888:3888
server.2=host-2:2888:3888
server.3=host-3:2888:3888

But, in host-1, I config host-1 resolve to 127.0.0.1 in /etc/hosts:

127.0.0.1   localhost host-1

which may results other hosts can't communicate with it. Resolve host-1 to its real ip solved this problem.

Hope this can help.



回答5:

I had this problem too, and it turned out that I was telling zookeeper to connect to the wrong port. Have you verified that zookeeper is actually running on port 2181 on the dev machine?



回答6:

I just have the same situation as you and I have just fixed this problem.

my conf/zoo.cfg just like this:

server.1=10.194.236.32:2888:3888
server.2=10.194.236.33:2888:3888
server.3=10.208.177.15:2888:3888
server.4=10.210.154.23:2888:3888
server.5=10.210.154.22:2888:3888

then i set data/myid file content like this:

1      //at host  10.194.236.32
2      //at host  10.194.236.33
3      //at host  10.208.177.15
4      //at host  10.210.154.23
5      //at host  10.210.154.22

finally restart zookeeper



回答7:

I was having the same error when I was trying to connect my broker with my Zookeeper ensemble using A records to point to Zookeeper IPs. The problem was in my zookeepers. My zookeepers were not able to bind to port 2181 because I was pointing my A records to public IP. This was preventing the zookeeper ensemble to choose a leader and communicate with each other. Pointing A records to private IP enabled the zookeeper ensemble to choose a leader and the cluster became active. After this when I tried to connect one of my brokers to the ensemble, it connected successfully.



回答8:

I had this problem too, and I found that I just need to restart zookeeper, then restart tomcat so my webapp connected nicely then



回答9:

I was able to start with zookeeper and kafka having 2 nodes each. I got the error because i had started zookeeper with ./zkServer.sh instead of the kafka wrapper bin/zookeeper-server-start.sh config/zookeeper.properties



回答10:

Make sure all required services are running

Step 1 : Check if hbase-master is running

sudo /etc/init.d/hbase-master status

if not, then start it sudo /etc/init.d/hbase-master start

Step 2 : Check if hbase-regionserver is running

sudo /etc/init.d/hbase-regionserver status

if not, then start it sudo /etc/init.d/hbase-regionserver start

Step 3 : Check if zookeeper-server is running

sudo /etc/init.d/zookeeper-server status

if not, then start it sudo /etc/init.d/zookeeper-server start


or simply run these 3 commands in a row.

sudo /etc/init.d/hbase-master restart
sudo /etc/init.d/hbase-regionserver restart
sudo /etc/init.d/zookeeper-server restart

after that don't forget to check the status

sudo /etc/init.d/hbase-master status
sudo /etc/init.d/hbase-regionserver status
sudo /etc/init.d/zookeeper-server status

You might find that zookeeper is still not running: then you can run the zookeeper

sudo /usr/lib/zookeeper/bin/zkServer.sh stop
sudo /usr/lib/zookeeper/bin/zkServer.sh start

after that again check the status and make sure its running

sudo /etc/init.d/zookeeper-server status

This should work.



回答11:

I start standalone instance in my machine, and encounter the same problem. Finally, I change from ip "127.0.0.1" to "localhost" and the problem is gone.



回答12:

This can happen if there are too many open connections.

Try increasing the maxClientCnxns setting.

From documentation:

maxClientCnxns (No Java system property)

Limits the number of concurrent connections (at the socket level) that a single client, identified by IP address, may make to a single member of the ZooKeeper ensemble. This is used to prevent certain classes of DoS attacks, including file descriptor exhaustion. Setting this to 0 or omitting it entirely removes the limit on concurrent connections.

You can edit settings in the config file. Most likely it can be found at /etc/zookeeper/conf/zoo.cfg.

In modern ZooKeeper versions default value is 60. You can increase it by adding the maxClientCnxns=4096 line to the end of the config file.



回答13:

I also ran into this problem last week and have managed to fix this now. I got the idea to resolve this one from the response shared by @gukoff.

My requirement and situation was slightly different from the ones shared so far but the issue was fundamentally the same so I thought of sharing it on this thread.

I was actually trying to query zookeeper quorum (after every 30 seconds) for some information from my application and was using the Curator Framework for this purpose (the methods available in LeaderLatch class). So, essentially I was starting up a CuratorFramework client and supplying this to LeaderLatch object.

Only after I ran into the error mentioned in this thread - I realised that I did not close the zookeeper client connection(s) established in my applications. The maxClientCnxns property had the value of 60 and as soon as the number of connections (all of them were stale connections) touched 60, my application started complaining with this error.

I found out about the number of open connections by:

  1. Checking the zookeeper logs, where there were warning messages stating "Too many connections from {IP address of the host}"

  2. Running the following netstat command from the same host mentioned in the above logs where my application was running:

netstat -no | grep :2181 | wc -l

Note: The 2181 port is the default for zookeeper supplied as a parameter in grep to match the zookeeper connections.

To fix this, I cleared up all of those stale connections manually and then added the code for closing the zookeeper client connections gracefully in my application.

I hope this helps!



回答14:

Also check the local firewall, service firewalld status

If it's running, just stop it service firewalld stop

And then give it a try.



回答15:

Check the zookeeper logs (/var/log/zookeeper). It looks like a connection is established, which should mean there is a record of it.

I had the same situation and it was because a process opened connections and failed to close them. This eventually exceeded the per-host connection limit and my logs were overflowing with

2016-08-03 15:21:13,201 [myid:] - WARN  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@188] - Too many connections from /172.31.38.64 - max is 50

Assuming zookeeper is on the usual port, you could do a check for that with:

lsof -i -P | grep 2181


回答16:

I encountered same problem ,too. In my case the problem is about iptables rules.

To communicate with zookeeper node, 2181 port must be accept incoming request, also for internal communication between zookeeper nodes 2888,3888 ports must be open for incoming request.

iptables -t nat -I PREROUTING -p tcp -s 10.0.0.0/24 --dport 2181 -j DNAT --to-destination serverIp:2181
iptables -t nat -I PREROUTING -p udp -s 10.0.0.0/24 --dport 2181 -j DNAT --to-destination serverIp:2181

iptables -t nat -I PREROUTING -p tcp -s 10.0.0.0/24 --dport 2888 -j DNAT --to-destination serverIp:2888
iptables -t nat -I PREROUTING -p udp -s 10.0.0.0/24 --dport 2888 -j DNAT --to-destination serverIp:2888

iptables -t nat -I PREROUTING -p tcp -s 10.0.0.0/24 --dport 3888 -j DNAT --to-destination serverIp:3888
iptables -t nat -I PREROUTING -p udp -s 10.0.0.0/24 --dport 3888 -j DNAT --to-destination serverIp:3888

sudo service iptables save


回答17:

This is a common issue if the Zookeeper server is not running or no longer running (i.e. it crashed after you started it).

So first, check that you have the Zookeeper server running. A simple way to check is grep the running processes:

# ps -ef | grep zookeeper

(run this a couple of times to see if the same process ID is still there. its possible it keep restarting with a new process ID. Alternatively you can use 'systemctl status zookeeper' if your Linux distro support systemd)

You should see the process running as a java process:

# ps -ef | grep zookeeper
root       492     0  0 00:01 pts/1    00:00:00 java -Dzookeeper.log.dir=. -Dzookeeper.root.logger=INFO,CONSOLE -cp /root/zookeeper-3.5.0-alpha/bin/../build/classes:/root/zookeeper-3.5.0-alpha/bin/../build/lib/*.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/slf4j-log4j12-1.7.5.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/slf4j-api-1.7.5.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/servlet-api-2.5-20081211.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/netty-3.7.0.Final.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/log4j-1.2.16.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/jline-2.11.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/jetty-util-6.1.26.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/jetty-6.1.26.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/javacc.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/jackson-mapper-asl-1.9.11.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/jackson-core-asl-1.9.11.jar:/root/zookeeper-3.5.0-alpha/bin/../lib/commons-cli-1.2.jar:/root/zookeeper-3.5.0-alpha/bin/../zookeeper-3.5.0-alpha.jar:/root/zookeeper-3.5.0-alpha/bin/../src/java/lib/*.jar:/root/zookeeper-3.5.0-alpha/bin/../conf: -Xmx1000m -Xmx1000m -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.local.only=false org.apache.zookeeper.server.quorum.QuorumPeerMain /root/zookeeper-3.5.0-alpha/bin/../conf/zoo.cfg

If its not there, then there's likely something in the zookeeper log file indicating the issue.

To find the zookeeper log file, you should first figure out where its configured for logging. In my case I have zookeeper installed under my root directory (not suggesting you install it there):

[root@centos6_zookeeper conf]# pwd
/root/zookeeper-3.5.0-alpha/conf

And you can find the log setting in this file:

[root@centos6_zookeeper conf]# grep "zookeeper.log" log4j.properties 
zookeeper.log.dir=/var/log
zookeeper.log.file=zookeeper.log
zookeeper.log.threshold=INFO
zookeeper.log.maxfilesize=256MB
zookeeper.log.maxbackupindex=20

So Zookeeper is configured to log under /var/log.

Then there's usually a zookeeper.log and/or zookeeper.out file which should indicate your startup error.



回答18:

Had the same error during setup on a 2 node cluster. I discovered I had mixed up the contents of the myid file versus the server.id=HOST_IP:port entry.

Essentially, if you have two servers (SERVER1 and SERVER2) for which you have created "myid" files in dataDir for zookeeper as below

SERVER1 (myid)
1

SERVER2 (myid)
2

Ensure the entry in your zoo.cfg file corresponds for each of these i.e server.1 should use SERVER1 hostname and server.2 should use SERVER2 hostname followed by the port as below

SERVER1 (zoo.cfg)
... (other config omitted)
server.1=SERVER1:2888:3888
server.2=SERVER2:2888:3888

SERVER2 (zoo.cfg)
... (other config omitted)
server.1=SERVER1:2888:3888
server.2=SERVER2:2888:3888

Just to make sure, I also deleted the version-* folder in the dataDir then restarted Zookeeper to get it working.



回答19:

Unable to read additional data from server sessionid 0x0, likely server has closed socket, closing socket connection and attempting reconnect (org.apache.zookeeper.ClientCnxn)

I changed just the number of brokers in the zoo.cfg file and restart zookeeper and kafka service



回答20:

I also get the same error when i started my replicated zk, one of zkClient can not connect to localhost:2181, i checked the log file under apache-zookeeper-3.5.5-bin/logs directory, and found this:

2019-08-20 11:30:39,763 [myid:5] - WARN [QuorumPeermyid=5(secure=disabled):QuorumCnxManager@677] - Cannot open channel to 3 at election address /xxxx:3888 java.net.SocketTimeoutException: connect timed out at java.net.PlainSocketImpl.socketConnect(Native Method) at java.net.AbstractPlainSocketImpl.doConnect(AbstractPlainSocketImpl.java:350) at java.net.AbstractPlainSocketImpl.connectToAddress(AbstractPlainSocketImpl.java:206) at java.net.AbstractPlainSocketImpl.connect(AbstractPlainSocketImpl.java:188) at java.net.SocksSocketImpl.connect(SocksSocketImpl.java:392) at java.net.Socket.connect(Socket.java:589) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectOne(QuorumCnxManager.java:648) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectOne(QuorumCnxManager.java:705) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectAll(QuorumCnxManager.java:733) at org.apache.zookeeper.server.quorum.FastLeaderElection.lookForLeader(FastLeaderElection.java:910) at org.apache.zookeeper.server.quorum.QuorumPeer.run(QuorumPeer.java:1247) 2019-08-20 11:30:44,768 [myid:5] - WARN [QuorumPeermyid=5(secure=disabled):QuorumCnxManager@677] - Cannot open channel to 4 at election address /xxxxxx:3888 java.net.SocketTimeoutException: connect timed out at java.net.PlainSocketImpl.socketConnect(Native Method) at java.net.AbstractPlainSocketImpl.doConnect(AbstractPlainSocketImpl.java:350) at java.net.AbstractPlainSocketImpl.connectToAddress(AbstractPlainSocketImpl.java:206) at java.net.AbstractPlainSocketImpl.connect(AbstractPlainSocketImpl.java:188) at java.net.SocksSocketImpl.connect(SocksSocketImpl.java:392) at java.net.Socket.connect(Socket.java:589) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectOne(QuorumCnxManager.java:648) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectOne(QuorumCnxManager.java:705) at org.apache.zookeeper.server.quorum.QuorumCnxManager.connectAll(QuorumCnxManager.java:733) at org.apache.zookeeper.server.quorum.FastLeaderElection.lookForLeader(FastLeaderElection.java:910) at org.apache.zookeeper.server.quorum.QuorumPeer.run(QuorumPeer.java:1247) 2019-08-20 11:30:44,769 [myid:5] - INFO [QuorumPeermyid=5(secure=disabled):FastLeaderElection@919] - Notification time out: 51200

that means this zk server can not connect to other servers, and i found this server ping other servers fail, and after remove this server from the replica, the problem is solved.

hope this will be helpful.



回答21:

This can happen despite the ZooKeeper servers being up and running and the socket open and accepting connections, if one or more of the ZooKeeper disks are out of space. This can easily happen if the old ZK snapshot and log files are never cleaned up:

The ZooKeeper server creates snapshot and log files, but never deletes them. The retention policy of the data and log files is implemented outside of the ZooKeeper server. The server itself only needs the latest complete fuzzy snapshot, all log files following it, and the last log file preceding it. The latter requirement is necessary to include updates which happened after this snapshot was started but went into the existing log file at that time. This is possible because snapshotting and rolling over of logs proceed somewhat independently in ZooKeeper. See the maintenance section in this document for more details on setting a retention policy and maintenance of ZooKeeper storage.

There is a maintenance job that can be run to clean up old snapshot and log files: See https://zookeeper.apache.org/doc/r3.4.12/zookeeperAdmin.html#sc_maintenance.



回答22:

leave only one entry for your host IP in /etc/hosts file, it resolved.



回答23:

Just now I solved the same question and post a blog.

In brief, if xx's zoo.cfg like:

server.1=xx:2888:3888
server.2=yy:2888:3888
server.3=zz:2888:3888

then xx's myid=1 is must