+ Start a Discussion

What is the recommended strategy for dealing with API network connectivity issues?

We have been experiencing "outages" on API calls. The API server (na1-api.salesforce.com) refuses to accept new SSL connections.

We have incorporated a retry strategy to retry the API call until it succeeds; with back-off time incrementing 100 milli-seconds each time. We expected a low number of retries before it succeeds.

However, that is not the case!
After the initial failure, our retry strategy kicks in as expected, but the recovery can take anywhere from 30 seconds to over a minute.

What is going on?? Is there a better retry/recovery strategy?
We cannot afford to have a thread waiting around for over a minute retrying an external SOAP call. Must we create our own job queue, to deal with possible "outages" ?

Another interesting data point.. It appears that these outages appear near the hour mark. Once at 2am, another time at 4pm (Eastern).
Do you have an exact time for that 4PM Eastern hiccup? I looked at my logs and I see a slight responce time increase in the 3:50-4:00 sample range, but nothing stastically significant. None of my threads exceeded their event windows during this period either (15 seconds).
From 3:58pm EST on 11/21/2005 to 4:00pm.
I have the Ethereal dump from two different locations running a test program.
I can upload/send the dump to you, if you like.