NullPointerException in worker when the leader of master was switching.

[2748...@qq.com]

Hi all,

   My environmen: centos6.6,spark 1.4.0,Tachyon 0.6.4,Gluster 3.4.2

   Master：node62->node52

   The logs in worker:

 2015-08-26 08:29:03,669 INFO WORKER_LOGGER (LeaderInquireClient.java:getMasterAddress) - Master addresses:[node62:19998, node52:19998]

2015-08-26 08:29:08,679 INFO WORKER_LOGGER (MasterClient.java:connect) - Tachyon client (version 0.6.4) is trying to connect master @ node52:19998
2015-08-26 08:29:08,683 ERROR WORKER_LOGGER (TachyonWorker.java:main) - Uncaught exception terminating worker
java.lang.NullPointerException
at tachyon.util.NetworkUtils.getFqdnHost(NetworkUtils.java:143)
at tachyon.master.MasterClient.connect(MasterClient.java:183)
at tachyon.master.MasterClient.worker_register(MasterClient.java:817)
at tachyon.worker.WorkerStorage.register(WorkerStorage.java:780)
at tachyon.worker.WorkerStorage.initialize(WorkerStorage.java:336)
at tachyon.worker.TachyonWorker.<init>(TachyonWorker.java:206)
at tachyon.worker.TachyonWorker.createWorker(TachyonWorker.java:90)
at tachyon.worker.TachyonWorker.main(TachyonWorker.java:126)

[Haoyuan Li]

Is the problem solved?

Best,

Haoyuan

--
You received this message because you are subscribed to the Google Groups "Tachyon Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tachyon-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[2748...@qq.com]

I don't find the reason yet,but the worker will recover itself when the master changes again.

Thanks,

Mingkai

在 2015年9月1日星期二 UTC+8下午7:33:34，Haoyuan Li写道：

[Haoyuan Li]

Why would the master change? Multiple master and one failed?

Thanks.

Haoyuan

--

Haoyuan Li

CEO, Tachyon Nexus

[野火燎原]

Our fault tolerant tachyon cluster have two master nodes and five slave nodes.Our spark cluster connect to the cluster using a public IP for the multiple master.The problem occurred when I testing the tachyon cluster.

Thanks.

Mingkai

------------------ 原始邮件 ------------------

发件人: "Haoyuan Li";<haoyu...@gmail.com>;

发送时间: 2015年9月2日(星期三) 晚上11:13

收件人: "野火燎原"<2748...@qq.com>;

抄送: "Tachyon Users"<tachyo...@googlegroups.com>;

主题: Re: NullPointerException in worker when the leader of master was switching.

[Haoyuan Li]

Are you running Zookeeper?

Haoyuan

[野火燎原]

Yes,I have three Zookeeper nodes.

My environmen: 

1.tachyon master: node52,node62

2.tachyon slave: node62,node60,node63,node104,node106

3.zookeeper:node62,node104,node106

Thanks,

Mingkai

------------------ 原始邮件 ------------------

发件人: "Haoyuan Li";<haoyu...@gmail.com>;

发送时间: 2015年9月7日(星期一) 中午11:34

[Haoyuan Li]

Mingkai,

Is this resolved? It will be great to give Tachyon 0.8.2 a try. 0.6.4 is very old. From Github, it's more than 6000 commits different from 0.8.2. :)

Looking forward to hearing from you.

Best,

Haoyuan

[野火燎原]

Haoyuan,

    

    I will have a try sometimes later, because my testing environment were occupied by my colleagues. I want to upgrade Tachyon 0.6.4 to 0.8.2 in my production environment because of some bugs in 0.6.4 , do you think it is a good 

idea?

Thanks,

Mingkai

------------------ 原始邮件 ------------------

发件人: "Haoyuan Li";<haoyu...@gmail.com>;

发送时间: 2015年11月16日(星期一) 凌晨2:52

[Haoyuan Li]

Definitely. 0.8.2 has many new features and is also much stable than 0.6.4.

Cheers,

Haoyuan

[2748...@qq.com]

Hi Haoyuan,

   

    I have test 0.8.2,the problem have solved.Thank you very much.

Thanks,

Mingkai

在 2015年11月19日星期四 UTC+8下午3:17:34，Haoyuan Li写道：

[Gene Pang]

Thanks for the confirmation!

-Gene

[Haoyuan Li]

Great!

On Wed, Dec 9, 2015 at 7:09 AM, Gene Pang <gene...@gmail.com> wrote:

Thanks for the confirmation!

-Gene

--
You received this message because you are subscribed to the Google Groups "Tachyon Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tachyon-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.