LAE

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: credoblado

Hi,

We have been using Checkpoints for some time now and have recently upgraded to LAE 4.5.2 (desktop version).

Our graph has 29 Checkpoints and at the end of the last run we noted only 18 Checkpoints had executed and have a "green tick" above the node. For the Checkpoints that have not "run" we can see the nodes before and after it have executed successfully.
Note that when the graph was run, there were a few stops and starts due to database execution issues.

We have only noticed this behaviour since the updgrade.

Why would Checkpoints not run?
Does the Checkpoint node run immediately after the previous node is run or does it wait?

Thanks,
Christina

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rboccuzzi

This could occur for a couple of different reasons, depending on your interaction with the graph. I don't believe that behavior has changed in 4.5.x at all.

If you want to understand more details, let me first explain checkpoints a bit more. A checkpoint node is made up internally of two nodes, and at any given time, only one of them is enabled, depending on whether you are configured to write the checkpoint or use the checkpoint. As you change the setting from writing the checkpoint to using it, the nodes inside flip in their enabled state.

One way that a checkpoint could be showing no tick but the nodes after have it, is if the checkpoint state was toggled after the run. This would cause the internal node that executed the checkpoint to be disabled, and the other node that has not yet run to be enabled.

Another way the checkpoint could be in that state depends on how you caused the following downstream nodes to execute. If the checkpoint node has Use Checkpoint set to false, and you don't actually select the checkpoint node but select nodes downstream and execute them (assuming you have execute predecessors enabled as a setting), it will only execute the preceeding nodes it HAS to execute to make your current selected node execute. And the checkpoint node in that configuration will not force the checkpoint to be written, which would cause the tick to not be there.

Please let me know if you are seeing this behavior and it can't be explained by the above. It may be I missed a legitimate behavior, or there is a new issue we can address.

Cheers
Rich

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: Edgois

Rob,

You initial explanation of the toggling of the checkpoint makes sense, however we know that the run in question would not be impacted by this issue.

You note that if the parameter is set to false, nodes before and after the checkpoints will run however the checkpoint will not write unless you select it run.

So my question is, what should the parameter be set to in order for the Checkpoint to automatically write?

Cheers
Emmanuel

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: stonysmith

One way to ensure that Checkpoints are written is to use the CLOCK feature.

Connect the OUTclock of the Checkpoint to the INclock of some downstream node. That way, the checkpoint must execute successfully before any downstream nodes will run.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: credoblado

Thanks for the tip. We have tried clocking our composites together which contain the checkpoint nodes and they run successfully.