E3SM-Project
EAMxx: make skip_t0_output true by default
We currently special-case this to be true for ERS/ERP tests, but we really oughta make it true by default
This would have the side effect of changing the first output file name. E.g., if the run starts at 2000-01-01-00000, and output is INSTANT every 5 min, the first file will have timestamp 2000-01-01-00600. I assume that's acceptable?
Yeah, that's fine. The "conventional" (e.g., EAM) behavior is exactly this:
if the run starts at 2000-01-01-00000, and output is INSTANT every 5 min, the first file will have timestamp 2000-01-01-00600
So, if someone wants something different, they should toggle the flag...
Do you remember if people specifically requested something different back in the day? We could probably review when support for this was added and see what the thinking was at the time?
Do you remember if people specifically requested something different back in the day? We could probably review when support for this was added and see what the thinking was at the time?
I don't think anybody requested anything different. I think that's always been the behavior. From a recent message (maybe on slack? by Rob maybe?) I found out EAM puts t0 output in its own file. In EAMxx, we just stuffed it in the same output as following instants. Maybe that generated some discrepancy and confusion.
As for the file name, users can also ask for 1 file per month (if that's not too big). We could even introduce "1 file per day", for high frequency output, if that helped. This would take away the annoyint "-00600" part from the filename.
I found out EAM puts t0 output in its own file.
Hmmm. I don't think this is the default behavior... I've never seen it!
I thought Rob said something like that, but it's quite possible I misread/misinterpreted.
@bartgol and @mahf708, when looking at the ne256 run from @hassanbeydoun , I noticed an unexpected behavior of the instantaneous output. Taking 3hi for example, from 1994-10-01 to 1995-09-20, the 3hour outputs start from 0h of the day, while from 1995-04-01 to end of the simulation, there is another set of 3hourly outputs starting from 3h of the day. It seams to be a restart problem, and the closest issue i can find in this one. Could you please help explain? Is it standard behavior to have this staggered time for instantaneous high frequency data upon restart?
Some questions to better diagnose:
- was the output OK from 1994-10-01 all the way to 1995-03-31 and bad from 1995-04-01?
- what was the restart frequency?
- did you specify any max file capacity in the output yaml files?
It seams to be a restart problem, and the closest issue i can find in this one.
I think it's the same general issue (and my motivation to open it in the first place)