v3.0

version 3 0 4 goals improve propagation improve log readability improved storage management provide more control to the operators upgrade the upgrade does not require any changes in configuration when upgrading from 1 1 this release has been available on the preview tag for several weeks the same release is now available on the latest tag the image path should be set to the following public ecr aws/cornucopias/nodes/pool server\ latest once it is time to upgrade, the following command will pull an updated image and restart the pool server using the new image docker compose pull docker compose up d the version of the pool server can be verified by looking at the first few lines of the log output \[2026 04 01t14 57 01z info filenodes pool server] pool server version 3 0 4 changes the propagation is now running slow and steady pool server will pull download small groupings of chunks then activate them (making them available to others) this will loop until all of the chunks are downloaded registrations will not be skipped when downloading a long time the log output should be run at the info level which is the default log level the logs should provide a clear story of what is happening errors should be understandable actions should be clear timings should be available an existing environmental variable should be understood filenodes pool max cache size is a variable that should be set to the max available storage the default is 100g the intent is that the pool will never exceed this storage amount more or less storage is ok because we will use what you give us the new normal is that we will keep this as full as possible a new environmental variable filenodes pool concurrent downloads will determine how fast you download the cache this is not a race, so please do not go crazy with this one for the sake of slow and steady, defaulting to 1 should be adequate for 99% of pools the last documented environmental variable is filenodes pool download batch size which will determine how many chunks to download before activating and making them available to other pool servers to download before doing any work, the pool server much know where it is, so it does requests to all of the system pools to measure latencies we have reduced the number of latency requests per system pool from 5 to 3 this will make the check a little quicker fixed a bug that when a system pool is down, the pool server will choose the region of the system pool that was down sample configurations note no changes are necessary for the upgrade these are just examples on how one may want to update their configurations for modified behavior default services pool server image public ecr aws/cornucopias/nodes/pool server\ latest ports \ "8001 8001" restart unless stopped environment filenodes pool access key "paste access key here" filenodes pool public port "8001" filenodes pool max cache size "100" filenodes pool concurrent downloads "1" volumes \ d \cornpoolserver\cache /cache lots of storage services pool server image public ecr aws/cornucopias/nodes/pool server\ latest ports \ "8001 8001" restart unless stopped environment filenodes pool access key "paste access key here" filenodes pool public port "8001" filenodes pool max cache size "1000" filenodes pool concurrent downloads "2" volumes \ d \cornpoolserver\cache /cache change log 3 0 2 failure retry delay for get cache files() increased from fixed 60 seconds to a random jitter between 10 20 minutes to avoid thundering herd on api timeouts registration log now shows duration until expiration in minutes and seconds (e g "58m 47s") to help confirm jitter in registration ttls 3 0 3 jwks verifying keys are now refreshed every registration cycle instead of once at startup, so key rotations take effect without restarting the pool servers 3 0 4 pass source pool's registration id (instead of client's own) to the downloader, fixing misleading "from registration" log messages and enabling correct skip tracking of unreachable source pools upgrade expectations it is expected that all users upgrade to 3 0 by june 1, 2026