Priam

New Feature

• (#821): Backport Configuring JVM Options for Cassandra using jvm.options if your Cassandra supports jvm.options, you can enable it by setting Priam.jvm.options.supported, then the options are the same as 3.11

Bug fix

• (#822): Use replace_address instead of replace_address_first_boot. replace_address always try to bootstrap Cassandra in replace mode even when the previous bootstrap is successful. replace_address_first_boot tries to bootstrap normally if the node already bootstrapped successfully.

Priam will check Cassandra gossip information while grabbing pre-assigned token to decide if it should start Cassandra in bootstrap mode or in replace mode. Moved token owner inferring logic based on Cassandra gossip into a util class. Refactored InstanceIdentity.init() method. Fixed DeadTokenRetriever to use the IP from token database when minimum number of instances are not reachable.

1. Priam will check Cassandra gossip information while grabbing pre-assigned token to decide if it should start Cassandra in bootstrap mode or in replace mode.
2. At most 3 random nodes are used to get the gossip information.
3. Moved token owner inferring logic based on Cassandra gossip into a util class.
4. Refactored InstanceIdentity.init() method.

• Update the backup service based on configuration changes.
• Expose the list of files from backups as API call.
• Run TTL for backup based on a simple timer to avoid S3 delete API call throttle.
• API to clear the local filesystem cache.
• Bug fix: Increment backup failure metric when no backup is found.
• Bug fix: No backup verification job during restore.

• Update the backup service based on configuration changes.
• Expose the list of files from backups as API call.
• Run TTL for backup based on a simple timer to avoid S3 delete API call throttle.
• API to clear the local filesystem cache.
• Bug fix: Increment backup failure metric when no backup is found.
• Bug fix: No backup verification job during restore.

• Write-thru cache in AbstractFileSystem.
• Take care of issue - C* snapshot w.r.t. filesystem is not "sync" in nature.

• Write-thru cache in AbstractFileSystem.
• Take care of issue - C* snapshot w.r.t. filesystem is not "sync" in nature.

• Fix for forgotten file
• Use older API for prefix filtering (backup TTL), if prefix is available.
• Send notifications only when we upload a file.

• Fix for forgotten file
• Use older API for prefix filtering (backup TTL), if prefix is available.
• Send notifications only when we upload a file.

BackupServlet had an API call of backup status which was producing output which was not JSON.

BackupServlet had an API call of backup status which was producing output which was not JSON.

(#784): BackupVerificationService
(#781): Put a cache for the getObjectExist API call to S3. This will help keep the cost of this call at bay.
(#781): Put a rate limiter for getObjectExist API call to S3 so we can limit the no. of calls.
(#783): Provide an override method to force Priam to replace a particular IP.

(#773): BackupVerificationService
(#779): Put a cache for the getObjectExist API call to S3. This will help keep the cost of this call at bay.
(#779): Put a rate limiter for getObjectExist API call to S3 so we can limit the no. of calls.
(#782): Provide an override method to force Priam to replace a particular IP.

S3 may decide to slow down and throw an error. Best not to do s3 object check (API) if it is not required.