We keep facing data synchronisation issues on several instances of our deployments of the three-components-based application (Fuseki, PrezAPI, Prez UI). Two of these instances are public at https://terra-vocabulary.org/portal/ld and https://terra-vocabulary.brgm-rec.fr/portal/ld. Respectively, each of their three components run on different docker containers hosted on a single virtual machine. These containers also bear different host names that are resolved by a root Apache server we call "resolver" on top of the PrezUI reverse proxy configuration suggested on #444 and in charge of rewriting URIs to their matching Prez-generated URLs.
Parts of the Nuxt-backed communications inbetween PrezAPI and PrezUI seem to be muted, causing desynchronisation actual data in Fuseki, what is replicated in PrezAPI's Pyoxigraph, and what's actully shown through PrezUI. As an example, the Fuseki server behind the -rec instance currently contains several catalogs (SHACL-validated to ensure their Prezentability), yet a different amount ascends back up through PrezUI.
From our infrastructure perspective it would look like PrezAPI is unable to notify PrezUI it shall make whatever query it needs to fetch the latest data pushed to Fuseki. An awkward solution we've picked so far is to restart every container (compose down then up), but it seems a more sustainable option is to maintain an applicative resolver, likely more costly that Apache redirects.
We keep facing data synchronisation issues on several instances of our deployments of the three-components-based application (Fuseki, PrezAPI, Prez UI). Two of these instances are public at https://terra-vocabulary.org/portal/ld and https://terra-vocabulary.brgm-rec.fr/portal/ld. Respectively, each of their three components run on different docker containers hosted on a single virtual machine. These containers also bear different host names that are resolved by a root Apache server we call "resolver" on top of the PrezUI reverse proxy configuration suggested on #444 and in charge of rewriting URIs to their matching Prez-generated URLs.
Parts of the Nuxt-backed communications inbetween PrezAPI and PrezUI seem to be muted, causing desynchronisation actual data in Fuseki, what is replicated in PrezAPI's Pyoxigraph, and what's actully shown through PrezUI. As an example, the Fuseki server behind the -rec instance currently contains several catalogs (SHACL-validated to ensure their Prezentability), yet a different amount ascends back up through PrezUI.
From our infrastructure perspective it would look like PrezAPI is unable to notify PrezUI it shall make whatever query it needs to fetch the latest data pushed to Fuseki. An awkward solution we've picked so far is to restart every container (compose down then up), but it seems a more sustainable option is to maintain an applicative resolver, likely more costly that Apache redirects.