giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jaeho Shin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (GIRAPH-904) Giraph can hang when hostnames include uppercase letters
Date Fri, 30 May 2014 21:38:01 GMT

    [ https://issues.apache.org/jira/browse/GIRAPH-904?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14014247#comment-14014247
] 

Jaeho Shin commented on GIRAPH-904:
-----------------------------------

The exact code where Giraph hangs is {{BspServiceMaster#barrierOnWorkerList()}}.  It keeps
printing log lines starting with {{"barrierOnWorkerList: Waiting on "}} with hostnames that
has definitely finished, e.g., when launched with a single worker, it waits for itself after
finishing the loading.

On a second look, I think a more precise fix would be doing the lowercase normalization in
that function when computing the {{hostnameIdList}} rather than doing it in {{GraphConfiguration#getLocalHostname()}}.
 The case difference between that and the hostname returned by {{TaskInfo#getHostnameId()}}
is causing the set difference computation in the while loop to fail.

> Giraph can hang when hostnames include uppercase letters
> --------------------------------------------------------
>
>                 Key: GIRAPH-904
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-904
>             Project: Giraph
>          Issue Type: Bug
>          Components: bsp, conf and scripts, zookeeper
>    Affects Versions: 1.1.0
>            Reporter: Jaeho Shin
>             Fix For: 1.1.0
>
>         Attachments: GIRAPH-904.patch
>
>
> We found that Giraph jobs were consistently hanging if uppercase letters were included
in the DNS (or /etc/hosts) resolved hostnames ({{foo.stanford.edu}} vs. {{foo.Stanford.EDU}}
from our DNS).  Normalizing the hostnames to lower case from {{GiraphConfiguration#getLocalHostname()}}
fixed our problem.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message