I have two free-running jobs (u-eb430 and u-eb461) that have been running successfully from the start of 1979, but both failed in the atmos_main step with the following error:
???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 4
? Error from routine: STWORK
? Error message: STWORK: Number of fields exceeds reserved headers for unit 11
? Error from processor: 68
? Error number: 71
???
u-eb430 failed for during 19800401T0000Z and u-eb461 failed during 19800301T0000Z. I haven’t changed any settings in the rose suite since setting the runs off.
This error means that you’ve sent too many fields to a particular output stream. You’d need to search through the pe0 file to see what unit 11 is referencing, and then check your STASH output to see what is being sent there. It’s odd that it’s failed so far into the run. The file will tell you the usage (the *.pa* file is UPA etc.), and then in the definition of e.g. UPA it will tell you the pp file stream, e.g. pp0 etc.).
Do you have any output that is only starting after a certain period of time, or is the file not actually being reinitialised so it’s filling up?
There are 2 solutions to this, which could be tried separately or combined. You can increase the number of headers (the reserved_headers entry in the pp file stream definition), or you can change the reinitialisation time, e.g. change to daily files from monthly etc. However, I’d first want to know why it’s taken this long to fail before doing either or both of these.
Note that a header is a level in a field, so a full 3D field would typically be 85 headers for an L85 model output on model levels, although this may be more if it’s also got pseudolevels as well. Surface/single-level fields are just one header.
Thanks for your response. Its coming from the STASH output stream upn which I included in this suite to output daily mean data. The header for this stream is already set to 860000 and it currently has 22 daily mean DALLTH fields outputting to it. The reinit_unit is ‘Real Months’ reinit_start is 0, reinit_end is -1, and reinit_step is 1.
I don’t think I have any outputs that start after a certain time period, I’m not sure how I would have set this up. It’s occurring on STASH 50138 which is a new STASH I have added but have run in nudged suites with no issues before :
ERROR detected in routine STWORK
: no. of output fields (=*****) exceeds no. of reserved PP headers for unit 11
STWORK: Error when processing diagnostic section 50, item 138, code 4
???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 4
? Error from routine: STWORK
? Error message: STWORK: Number of fields exceeds reserved headers for unit 11
? Error from processor: 0
? Error number: 71
???
Additionally, it appears that neither of the new output streams I added (ap3 and apn) are on MASS. I have checked and I included them in the rose suite in the exact same way as I did for the nudged suite where they have output successfully:
[izzysang@mass-cli2 ~]$ moo ls moose:/crum/u-eb430
moose:/crum/u-eb430/ada.file
moose:/crum/u-eb430/apa.pp
moose:/crum/u-eb430/apb.pp
moose:/crum/u-eb430/apc.pp
moose:/crum/u-eb430/apd.pp
moose:/crum/u-eb430/ape.pp
moose:/crum/u-eb430/apf.pp
moose:/crum/u-eb430/apg.pp
moose:/crum/u-eb430/api.pp
moose:/crum/u-eb430/apj.pp
moose:/crum/u-eb430/apk.pp
moose:/crum/u-eb430/apm.pp
moose:/crum/u-eb430/aps.pp
[izzysang@mass-cli2 ~]$ moo ls moose:/crum/u-eb461
moose:/crum/u-eb461/ada.file
moose:/crum/u-eb461/apa.pp
moose:/crum/u-eb461/apb.pp
moose:/crum/u-eb461/apc.pp
moose:/crum/u-eb461/apd.pp
moose:/crum/u-eb461/ape.pp
moose:/crum/u-eb461/apf.pp
moose:/crum/u-eb461/apg.pp
moose:/crum/u-eb461/api.pp
moose:/crum/u-eb461/apj.pp
moose:/crum/u-eb461/apk.pp
moose:/crum/u-eb461/apm.pp
moose:/crum/u-eb461/aps.pp
The ‘Real month’ setting is only relevant to model runs using 365-day calendar (e.g. for Nudging). This might be why the stream is being ignored and writing to it causing issues.
Have you defined the new streams by hand, or the GUI? Adding via the GUI will perform basic checks on the settings (and expose only the relevant ones) to make sure these are viable.