giraph-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yu, Jaewook" <jaewook...@intel.com>
Subject RE: Running ConnectedComponents in a cluster.
Date Thu, 17 Apr 2014 16:02:20 GMT
Ghufran,

It looks like the graph loading is failing from your log:

14/04/17 16:12:31 INFO job.JobProgressTracker: Data from 3 workers - Loading data: 0 vertices
loaded, 0 vertex input splits loaded; 0 edges loaded, 0 edge input splits loaded; min free
memory on worker 2 - 141.96MB, average 142.66MB

If you have access to JobTracker web interface (port 50030) or you know where the log files
are located, take a look at the log for this failing job. That would be a good starting point
to debug the issue.

Thanks,
Jae

From: ghufran malik [mailto:ghufran1malik@gmail.com]
Sent: Thursday, April 17, 2014 8:28 AM
To: user@giraph.apache.org
Subject: Re: Running ConnectedComponents in a cluster.

I would appreciate it if you could lend me your assistance with another problem of mine.
I have an implementation of TriangleCounting algorithm that runs correctly on the smaller
dataset I used to test ConnectedComponents, but fails when trying to compute this larger dataset.
the map seems to fail and I do not know why. Full output is below.

14/04/17 16:12:30 INFO job.HaltApplicationUtils$DefaultHaltInstructionsWriter: writeHaltInstructions:
To halt after next superstep execute: 'bin/halt-application --zkServer ricotta.eecs.qmul.ac.uk:2181<http://ricotta.eecs.qmul.ac.uk:2181>
--zkNode /_hadoopBsp/job_1381849812331_2770/_haltComputation'
14/04/17 16:12:31 INFO mapreduce.Job: Running job: job_1381849812331_2770
14/04/17 16:12:31 INFO job.JobProgressTracker: Data from 3 workers - Loading data: 0 vertices
loaded, 0 vertex input splits loaded; 0 edges loaded, 0 edge input splits loaded; min free
memory on worker 2 - 141.96MB, average 142.66MB
14/04/17 16:12:32 INFO mapreduce.Job: Job job_1381849812331_2770 running in uber mode : false
14/04/17 16:12:32 INFO mapreduce.Job:  map 100% reduce 0%
14/04/17 16:12:36 INFO job.JobProgressTracker: Data from 3 workers - Loading data: 0 vertices
loaded, 0 vertex input splits loaded; 0 edges loaded, 0 edge input splits loaded; min free
memory on worker 2 - 141.96MB, average 142.66MB
14/04/17 16:12:41 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 1:
0 out of 378222 vertices computed; 0 out of 3 partitions computed; min free memory on worker
2 - 24.77MB, average 103.6MB
14/04/17 16:12:46 INFO job.JobProgressTracker: Data from 3 workers - Compute superstep 1:
0 out of 1134723 vertices computed; 0 out of 9 partitions computed; min free memory on worker
1 - 22.5MB, average 23.36MB
14/04/17 16:12:48 INFO mapreduce.Job: Job job_1381849812331_2770 failed with state FAILED
due to: Task failed task_1381849812331_2770_m_000002
Job failed as tasks failed. failedMaps:1 failedReduces:0

14/04/17 16:12:48 INFO mapreduce.Job: Counters: 46
    File System Counters
        FILE: Number of bytes read=0
        FILE: Number of bytes written=143668
        FILE: Number of read operations=0
        FILE: Number of large read operations=0
        FILE: Number of write operations=0
        HDFS: Number of bytes read=37028489
        HDFS: Number of bytes written=0
        HDFS: Number of read operations=3
        HDFS: Number of large read operations=0
        HDFS: Number of write operations=0
    Job Counters
        Failed map tasks=1
        Launched map tasks=3
        Other local map tasks=3
        Total time spent by all maps in occupied slots (ms)=24219
        Total time spent by all reduces in occupied slots (ms)=0
    Map-Reduce Framework
        Map input records=2
        Map output records=0
        Input split bytes=88
        Spilled Records=0
        Failed Shuffles=0
        Merged Map outputs=0
        GC time elapsed (ms)=22209
        CPU time spent (ms)=77200
        Physical memory (bytes) snapshot=659660800
        Virtual memory (bytes) snapshot=1657229312
        Total committed heap usage (bytes)=372899840
    Giraph Stats
        Aggregate edges=0
        Aggregate finished vertices=0
        Aggregate sent message message bytes=0
        Aggregate sent messages=0
        Aggregate vertices=0
        Current master task partition=0
        Current workers=0
        Last checkpointed superstep=0
        Sent message bytes=0
        Sent messages=0
        Superstep=0
    Giraph Timers
        Initialize (ms)=0
        Setup (ms)=0
        Shutdown (ms)=0
        Total (ms)=0
    Zookeeper base path
        /_hadoopBsp/job_1381849812331_2770=0
    Zookeeper halt node
        /_hadoopBsp/job_1381849812331_2770/_haltComputation=0
    Zookeeper server:port
        ricotta.eecs.qmul.ac.uk:2181=0
    File Input Format Counters
        Bytes Read=0
    File Output Format Counters
        Bytes Written=0
Thanks,
Ghufran

On Thu, Apr 17, 2014 at 4:21 PM, ghufran malik <ghufran1malik@gmail.com<mailto:ghufran1malik@gmail.com>>
wrote:
oh woops! yes i meant i change it to an undirected format!

On Thu, Apr 17, 2014 at 4:11 PM, ghufran malik <ghufran1malik@gmail.com<mailto:ghufran1malik@gmail.com>>
wrote:
Hi Jae,
Thanks so much for pointing out that it wasn't directed. I made the changes and made a directed
graph and connected components now works :)
Thanks,
Ghufran

On Wed, Apr 16, 2014 at 7:31 PM, Yu, Jaewook <jaewook.yu@intel.com<mailto:jaewook.yu@intel.com>>
wrote:
Ghufran,

The Youtube community dataset (com-youtube.ungraph.txt.gz<https://snap.stanford.edu/data/bigdata/communities/com-youtube.ungraph.txt.gz>)
[1] is formatted as directed graph although the description says it’s undirected graph.
With some minor changes in your conversion program, you should be able to generated a proper
undirected adjacency list.

Hope this will help.

Thanks,
Jae

[1] https://snap.stanford.edu/data/com-Youtube.html

From: Yu, Jaewook [mailto:jaewook.yu@intel.com<mailto:jaewook.yu@intel.com>]
Sent: Wednesday, April 16, 2014 11:00 AM
To: user@giraph.apache.org<mailto:user@giraph.apache.org>
Subject: RE: Running ConnectedComponents in a cluster.

Hi Ghufran,

Have you verified the neighbors of each vertex actually exist? From your adjacency list, for
example, 278447 278447 532613, is the neighbor’s vertex id 532613 valid?

Thanks,
Jae


From: ghufran malik [mailto:ghufran1malik@gmail.com]
Sent: Wednesday, April 16, 2014 9:22 AM
To: user@giraph.apache.org<mailto:user@giraph.apache.org>
Subject: Running ConnectedComponents in a cluster.

Hi,
I have setup Giraph on my university cluster of computers (Giraph 1.1.0-SNAPSHOT-for-hadoop-2.0.0-cdh4.3.1).
I've successfully ran the connected components algorithm on a very small test dataset using
4 workers and it produced the expected output.

dataset:

vertex id, vertex value, neighbours....

0 0 1
1 1 0 2 3
2 2 1 3
3 3 1 2
output:
1    0
0    0
3    0
2    0

However when I tried to run this algorithm on a larger dataset (reformatted version of com-youtube.ungraph
from Stanford snap to match the IntIntNullTextVertexInputFormat) it successfully complets
but the incorrect output is produced. It seems to just output the vertex id with its orignal
value (its vertex id is its original value that i set).
A snippet of the dataset is provided:
vertex id, vertex value, neighbours....
.......
278447 278447 532613
278449 278449 305447 324115 414238
83899 83899 153460 172614 176613 211448
773749 773749 845366
773748 773748 960388
.......
output produced:
.............
73132    73132<tel:73132%C2%A0%C2%A0%C2%A0%2073132>
831308    831308
199788    199788
763644    763644
300572    300572
.............
there's not one vertex value that isn't the same as its original vertex ID.
The computation also stops after superstep 0 is done and goes no further, whereas on my smaller
data set completes 3 supersteps.
Does anyone have an idea to why this is?
Kind regards,
Ghufran




Mime
View raw message