Item | Description | ID |
|---|
1 | data-layer: Fixed incorrect handling of invalid values in control-plane monitoring. In unexpected cases in which a control-plane key-value object could mistakenly be stored with an invalid value, Lightbits would stop monitoring for specific key/prefix changes, which could cause the control plane to malfunction.
| LBM1-35501 |
2 | data-layer: Fixed an issue that could cause a node to hang when etcd becomes unresponsive.
| LBM1-35894 |
3 | data-layer: Fixed incorrect error handling when watching for changes in the control plain. Due to a missing retry mechanism, in some rare cases, Lightbits could potentially completely stop watching for changes - without raising any significant user event or attempting to restart the watch. This could potentially lead to service loss. Upgrading from 3.10.x, 3.11.x, and 3.12.1 is highly recommended.
| LBM1-36090 |
4 | duroslight: Fixed a bug that could cause slightly longer than needed rebuilds due to missed synchronization points - causing Lightbits to rebuild based on an earlier synchronization point.
| LBM1-35565 |
5 | lb-support: Did not bring kdumpinformation if kdumpwas not used on the system.
| LBM1-36260 LBM1-35753 |
6 | node-manager: A failure - while starting to run the internal services of the node-manager - could cause the node to set its status to active, although it cannot handle any change in the protection group.
| LBM1-35886 |
7 | node-manager: Improved the node-manager startup flow. Added previously missing events to different failure flows, and retry-mechanisms for transient network issues when loading different data-layer objects.
| LBM1-35749 |
8 | node-manager: Made TPM encryption handling more robust in the face of reusing servers in different clusters, by not relying on existing TPM sub-keys for encryption.
| LBM1-36027 |
9 | node-manager: Fixed a bug that caused a node to remain inactive if encryption was enabled using TPM, but then the TPM was mistakenly cleared on the node's server (and then the node-manager service was restarted).
| LBM1-36044 |
10 | node-manager: Fixed an issue where a failure while starting to run the internal services of the node-manager could cause the node to appear as active, when it is actually inactive.
| LBM1-35575 |
11 | node-manager: Increased the default LimitNOFILE the node-manager sets in the duroslight-*.service files from 65536 to 300,000. This is to prevent duroslight process failure due to "Too many open files" with a large number of connected clients. duroslight will gracefully reject new connections once the limit of 2030 clients is reached. Each client can have up to 128 I/O connections.
| LBM1-36478 |
12 | node-manager: Fixed an issue that can lead to a node becoming inactive under certain circumstances, where TPM access completely jams the go runtime scheduler, preventing any other go-routines spawned by the running service from running.
| LBM1-35340 |
13 | node-manager: Prevented the possibility of rejoining a cluster after a server has been deleted from the cluster.
| LBM1-35739 LBM1-36018 |
14 | packages/backend: The phase of TRIM object processing is now reflected in the powerup progress.
| LBM1-35611 |
15 | profile-generator: Fixed a potential hang on Alma9 kernels related to single-instance multi-numa deployment. The fix enables node-manager to reserve memory via huge pages and free the memory once GFTL completes allocating most of the system RAM – ensuring that every node has minimal RAM space left to avoid hangs due to endless node-reclaim on Alma 9 kernels. The default extraReservedRAM = 50% of the OS reserved (currently this is 4 GB).
| LBM1-35837 |
16 | upgrade-manager: Fixed an issue that caused the cluster last upgrade status to remain as "Upgrading" instead of "Failed", in case there was a failure in the cluster upgrade process.
| LBM1-35663 |
17 | userlbe: Resolved a rare issue where SSDs failing to return write completions for an extended period could cause the GFTL to crash, resulting in the node becoming inactive.
| LBM1-35598 |
18 | userlbe: Fixed a bug where - under some circumstances for graceful shutdown and recovery - Lightbits would use one core less than what was possible. This fix should slightly speed up graceful shutdown/recovery under those circumstances.
| LBM1-35733 |
19 | userlbe: Fixed an issue where the sorting algorithm used was not configured properly for large lists sorted during powerup, which caused graceful powerup to take significantly longer than it should.
| LBM1-35734 |
20 | userlbe: Fixed an issue where latency metrics were all bunched together in the smallest histogram bucket, making it harder to debug disk latency issues. Metrics are now correctly divided into buckets.
| LBM1-35882 |
