Globus Transfer

Server too busy to send notification emails?

Details

  • Type: Bug Bug
  • Status: Resolved Resolved
  • Priority: Major Major
  • Resolution: Fixed
  • Affects Version/s: None
  • Fix Version/s: Sprint 2010-07-23
  • Component/s: Backend
  • Labels:
    None
  • Ranking:
    Issue can not be ranked. 

Description

Per the findings in the third CEDPS data challenge (http://www.mcs.anl.gov/~childers/CDC3/):

After runs #3 and #4 some of the users did not receive transfer completion notification emails. It was a busy time for Globus.org, with 10,000 transfer groups for multiple users completing nearly at the same time.

Activity

Hide
Karl Pickett added a comment -
One issue that is not optimal with our email notification system is there is no queueing. We use sendEmail to try to open up a socket to gmail.com for SMTP relay and if that fails, there is no retry. We should probably use the local sendmail queue which handles retries. (using the sendmail command or similar).

This requires some sendmail (or other mta) study & hacking to forward to either gmail or a 'real' smtp server for globus.org.
Show
Karl Pickett added a comment - One issue that is not optimal with our email notification system is there is no queueing. We use sendEmail to try to open up a socket to gmail.com for SMTP relay and if that fails, there is no retry. We should probably use the local sendmail queue which handles retries. (using the sendmail command or similar). This requires some sendmail (or other mta) study & hacking to forward to either gmail or a 'real' smtp server for globus.org.
Hide
Karl Pickett added a comment -
I installed postfix as an mta and we use that now for queueing to gmail.com. This is the only issue that looked seriously broken for email notification.
Show
Karl Pickett added a comment - I installed postfix as an mta and we use that now for queueing to gmail.com. This is the only issue that looked seriously broken for email notification.

People

Vote (0)
Watch (0)

Dates

  • Created:
    Updated:
    Resolved: