Cache and disk saturation

[Léo Le Lonquer]

Hello,

One of our dataset is giving us troubles on our instance https://erddap.ifremer.fr.

Some requests on ArgoFloats dataset create disk overflow as the cache folder fills up.

Example :

2025-03-20T06:56:15+0100 [erddap.ifremer.fr:443] 502 GET "/erddap/tabledap/ArgoFloats.ncHeader" | qs:"?fileNumber%2Cdata_type%2Cformat_version%2Chandbook_version%2Creference_date_time%2Cdate_creation%2Cdate_update%2Cplatform_number%2Cproject_name%2Cpi_name%2Ccycle_number%2Cdirection%2Cdata_center%2Cdc_reference%2Cdata_state_indicator%2Cdata_mode%2Cplatform_type%2Cfloat_serial_no%2Cfirmware_version%2Cwmo_inst_type%2Ctime%2Ctime_qc%2Ctime_location%2Clatitude%2Clongitude%2Cposition_qc%2Cpositioning_system%2Cprofile_pres_qc%2Cprofile_temp_qc%2Cprofile_psal_qc%2Cvertical_sampling_scheme%2Cconfig_mission_number%2Cpres%2Cpres_qc%2Cpres_adjusted%2Cpres_adjusted_qc%2Cpres_adjusted_error%2Ctemp%2Ctemp_qc%2Ctemp_adjusted%2Ctemp_adjusted_qc%2Ctemp_adjusted_error%2Cpsal%2Cpsal_qc%2Cpsal_adjusted%2Cpsal_adjusted_qc%2Cpsal_adjusted_error%2Cdoxy%2Cdoxy_qc%2Ctemp_doxy%2Ctemp_doxy_qc%2Cmolar_doxy%2Cmolar_doxy_qc%2Cturbidity%2Cturbidity_qc%2Cchla%2Cchla_qc%2Cnitrate%2Cnitrate_qc&time%3E=1997-07-28T20%3A26%3A20Z&time%3C=2026-12-27T14%3A48%3A20Z" | d:60081617 us | p:HTTP/1.1 | ua:"python-requests/2.32.3" | ref:"-" | reqSz:1793 | respSz:6287 | uid:- | rid:Z9ut-4b2jicAAHOBRXwAAAMY | sid:- | ip:131.217.255.240 | cs:+ | ka:0

When the cache partition is completely full, the erddap instance crashes.

We tried to set cacheMinutes to 30 minutes but it doesn't stop the overlarding.

The total size of ArgoFloats is 68G.

Is there a way to prevent the instance from crashing ?

Why cacheMinutes isn't solving this issue ?

Regards

--

Léo Le Lonquer
Ingénieur Cloud pour la Science
Ifremer Centre Bretagne, 29280 Plouzané, France