Bug #1413

wget agent on clusters thinks path should be http* or ftp

Added by Mark Donohoe about 1 year ago. Updated about 1 year ago.

Status:Closed Start date:03/06/2012
Priority:Low Due date:
Assignee:Mary Laser % Done:

0%
Category:Wget agent
Target version:2.0.0
Rank: Tester:

Description

Using svn 5571 on the snape cluster, a file that was trying to be uploaded from server failed. This only happens on a cluster. I was trying to upload /Fedora/Fedora16/Fedora-16-source-CHECKSUM from snape (the server).

wget_agent failed with the message:

FATAL wget_agent.c.386: path /Fedora/Fedora16/Fedora-16-source-CHECKSUM is not http://, https://, or ftp://"

If the same file is in /home/markd it can be uploaded from server without any problems. Why does that path (/Fedora/Fedora16/ make the wget_agent think it needs http or ftp protocol?

History

#1
Updated by larry shi about 1 year ago

this defect only happened in cluster.
the　reason is because the wget_agent server can not find the file that you want to upload from server.
1. when you upload from server, the path you want to upload is on the scheduler/ui server.
2. but for the wget_agent, probably it is on the same server with the schdeduler/ui, but probably not.
3. if the scheduler server is different with the wget_agent server, failure occur.

for now, I have no solution to fix this issue.

I remeber that at the beginning of 2.0, we want to put the wget_agent in fossology core, wget_agent/scheduler/ui should be on the same server.

any comments?

#2
Updated by larry shi about 1 year ago

  • Status changed from New to Feedback

#3
Updated by Mark Donohoe about 1 year ago

Larry,

I have sent email to Bob on this, I understand what you are saying, I thought that situation would be fixed in 2.0.
Let's see what Bob thinks.

#4
Updated by Mary Laser about 1 year ago

  • Assignee changed from larry shi to Mary Laser
  • Priority changed from Normal to Low

As we strive to get to RC1, let's leave this at a normal priority. As a workaround, the cluster config can restrict Upload from Server to the web server.

Mary to get details from Alex and add to the cluster config docs.

#5
Updated by Alex Norton about 1 year ago

  • Status changed from Feedback to Resolved

This should be resolved with svn 5626.

Agent configurations now have a special option of LOCAL which will cause the scheduler to only run those agents on localhost. Added this option to wget so that it should only get run on localhost and shouldn't have this problem anymore.

#6
Updated by larry shi about 1 year ago

  • Status changed from Resolved to In Progress

scheduler can not work in svn 5626

#7
Updated by larry shi about 1 year ago

  • Status changed from In Progress to Resolved

tested with svn 5642, it works

#8
Updated by larry shi about 1 year ago

  • Status changed from Resolved to Closed

verified on snape cluster, svn 5724.

Also available in: Atom PDF