I had a docker installation with NC21 (from NC19) with traefik and collabora as extra container. After upgrade to latest collabora image collabora/code:6.4.10.1 I experienced issue with error “Failed to read document from storage. Please contact your storage server (cloud.example.com) administrator.” as described here
triying to figure out if this is a collabora issue I went back to older collabora version 6.4.0.6 (unfortunately I used “latest” tag initially so I don’t really know which one I was in use before) and found the issue changed slightly. The document doesn’t fail but takes 3-5 minutes to load. once loaded editing works smooth.
I don’t see no related errors/issues in the logs (at least not with default settings) in browser F12 tools I see some requests on starting document load and then nothing happens only bunch of sync requests are send to the server
this request seems to the one where it start working (Status 101 Switching Protocols)
Maybe collabora somehow doesn’t understand we should use https? Regarding the config: TLS is terminated on traefik reverse proxy - collabora and Nextcloud run with plain http…
I made a lot of tests and the outcome is the problem is definitely related to the 101 Switching Protocols. This happens as the connection changes from http to websocket. the request looks like
GET
wss://<<<my collabora fqdn>>>/lool/https%3A%2F%2F<<<my NC fqdn>>>%2Findex.php%2Fapps%2Frichdocuments%2Fwopi%2Ffiles%2F4776_oc52dthqts8g%3Faccess_token%3DDh09QWOnFDf42Vs9KWFZTCh0q6NSrDay%26access_token_ttl%3D0/ws?WOPISrc=https://<<<my NC fqdn>>>/index.php/apps/richdocuments/wopi/files/4776_oc52dthqts8g&compat=/ws
Old collabora version 6.4.0.6 at this point (after 3.68 min) starts loading the document (and works as expected), newer version 6.4.10.1 fails with the “failed to read document…” error (after 2.?? min). the time spend in this state is constant for each specific version, which makes me think it must be some timeout before the system makes another attempt and starts working.
I don’t see any useful output in the log of both NC and collabora, even on debug level. Has someone any idea what happens before/while switching to wss://, which component is responsible for this and how I can dig further into the issue?
