workers blocking when reading hdfs files

2 views
Skip to first unread message

Ahmet Uyar

unread,
Jul 28, 2020, 8:55:38 AM7/28/20
to Twister2
Hi guys,

I am running 240 twister2 workers, each of them is supposed to read a separate hdfs file with 1000 tweetIDs in them. Usually they all read the files without any issues. But, sometimes they just block and wait indefinitely.  

I am attaching jstack outputs from two such workers. 

thanks,

Ahmet
jstack-blocked-worker-5.txt
jstack-blocked-worker-155.txt

Ahmet Uyar

unread,
Jul 29, 2020, 6:52:12 AM7/29/20
to Twister2
Hi guys,

I think this issue is not related to hdfs file reading. I think it is related to some communication issue. 
The Main worker thread is waiting on the method: mpi.Request.testStatus(Native Method).
It also happens after all workers have finished reading the input files for deleteIDs. 

Ahmet

Ahmet Uyar

unread,
Jul 29, 2020, 10:09:05 AM7/29/20
to Twister2
Hi guys,

This issue also happens when partitioning tweedID-date pairs. 
Attached jstack output is from such a run. 
The job blocks when partitioning tweetID-date pairs. 

Ahmet
jstack-blocked-w2235.txt

Gregor von Laszewski

unread,
Jul 31, 2020, 12:15:23 PM7/31/20
to Twister2
Last year we talked about educational material for Twister2 that is not just a Manual.
The gaol was to have some material that would introduce the students to twister2 and have
practical assignments such as

a) installation
b) hello word
c) examples on to do xyz

Now that a lot of changes have been made is such material available?

Originally this was supposed to be developed in markdown so it could be integrated into bookmanager
And our lecture notes, but I do not think that was done. I think instead the manual pages have been developed.

Is this right?

Let me know if such educational material is or is not available.

Thanks

Gregor

Supun Kamburugamuve

unread,
Aug 3, 2020, 9:36:28 AM8/3/20
to Gregor von Laszewski, Twister2
The current document is in 


There are few guides on compiling twister2, in mac and Linux.


There is a tutorial on getting started. 


The website is generated from markdown files found in the source. 


It uses docusaurus to generate the website. The website is hosted in Github.


Let us know, what you think. We can create/rearrange material as needed.

Best,
Supun..

--
You received this message because you are subscribed to the Google Groups "Twister2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to twister2+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/twister2/F45B10D2-534C-49F1-B7DF-E9BE32C83107%40gmail.com.


--
Supun Kamburugamuve, PhD
Digital Science Center, Indiana University
Member, Apache Software Foundation; http://www.apache.org
E-mail: supun@apache.org;  Mobile: +1 812 219 2563


Reply all
Reply to author
Forward
0 new messages