For Storage Nodes
In Memory Used Serials
The storage node keeps the used serials list in memory instead of writing them into a SQLite DB. Serial numbers are used for double spend protection. To minimize memory usage we decreased the expire time for uplinks from 24 hours to 1 hour. Within 1 hour the uplink needs to start the transfer because otherwise, the storage nodes will reject the order. The transfer itself can still take longer than 1 hour. Orders are also still valid for 48 hours. After one hour the storage node can delete the used serials without any consequences. By default storage2.max-used-serials-size
is 1 MB.
If the maximum size is reached the node will start dropping random serial numbers. If you want to optimize your storage node you should take a look into monkit stats for delete_random_serial
On a restart, the storage node will drop all serial numbers. For a short time, a malicious uplink could submit an old order twice while the storage node would get paid only once. However, the uplink still needs 29 pieces to reconstruct the file. Other storage nodes will keep rejecting the malicious uplink request.
500 MB Node Selection Threshold
With the previous release, we enabled the node selection cache on all satellites. The cache is refreshed every 3 minutes and as a result, nodes can still be selected for upload for up to 3 minutes after the node signaled that it’s nearly full. Nodes previously sent this signal when there was less than 100MB of space left, but this was insufficient and some nodes filled up that 100MB within the 3-minute delay. We raised it to 500MB to compensate for this.
Storage Node Payout History Fixed
We fixed the payout data on the satellite side. The storage node payment dashboard should show correct payout data from day 1 for all satellites.
There was also a problem with the download traffic accounting on saltlake. The accounting was too slow compared to the number of orders the satellite had to process. In the beginning, Saltlake was falling behind with the accounting only for a few especially bigger storage nodes. For the majority of the smaller storage nodes, saltlake was still able to process all orders. Over time the number of affected storage nodes increased slowly. We fixed it in the first week of June and are currently moving the data into the May payout.
Storage Node Dashboard
We added total earning to the storage node dashboard and also a diagram that shows used space, free space, and garbage. We are aware that the values shown in the storage node CLI are inconsistent with the WebUI dashboard. The CLI dashboard is showing piece size + metadata but the WebUI dashboard is showing piece size only. This will be fixed with the following release.
Suspension Mode Audit Score
Last release we added unknownAlpha
and unknownBeta
values to the storage node dashboard API. Now unknownScore
is available as well.
For Customers
Bucket Value Attribution
For open-source projects, we offer an open source partner program. Open source partners are getting a special partner ID and a short name from us. The short name can be used for a feature that we call bucket value attribution. More information on how to enable this feature can be found here: https://documentation.tardigrade.io/how-tos/configure-tools-for-the-partner-program