View Single Post
Old 2021-06-30, 22:32   #115
swellman
 
swellman's Avatar
 
Jun 2012

27·52 Posts
Default

Quote:
Originally Posted by Xyzzy View Post
We restarted it using the default target density. We didn't keep the log.

We tend to start jobs as soon as there are enough relations to build a matrix, usually before all of the pending relations have been returned. Could this maybe be the problem?

In general I wait until the Estimated Pending Relations < 1M before downloading the dataset, unless the job is clearly oversieved. But that’s what I do, not necessarily what is optimal. I suspect this job was undersieved even after all the relations had trickled in. Adding more relations may be the path to success.

But I’ve never seen a corrupt state error message. When the requested TD is too large or not enough relations are present msieve will demand more rels, but never a message with the word corrupt.
swellman is online now   Reply With Quote