You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Milad Fatenejad <ic...@gmail.com> on 2014/08/21 18:24:13 UTC

Working with tuples that may take a long time to process

Hello:

I am building a topology where the majority of my tuples will be fully
processed relatively quickly (within a second or so). However, I will
occasionally have tuples that require a long time to fully process (in the
absolute worst case, up to half an hour). My question is, how do I handle
these long running tuples most effectively since I have to choose a single
timeout for my topology (i.e. TOPOLOGY_MESSAGE_TIMEOUT_SECS). Here are my
current solutions:

1. Make the topology timeout really long (bad for a lot of reasons)
2. Create a separate topology for long running tasks that has a longer
timeout, but that seems like it introduces unnecessary redundancy.

Does anyone have any best-practices for dealing with this situation?

Also, I was wondering if there is some kind of "in-progress" signal I could
send to let storm know that even though a tuple is taking a long time, it
is still being processed...

Thank you!
Milad