https://docs.splunk.com/Documentation/Splunk/9.4.0/ReleaseNotes/MeetSplunk#What.27s_New_in_9.4
Why New Splunk TcpOutput Persistent Queue?
- Scheduled no connectivity for extended period but need to resume data transmission once connection is back up. Assuming there is enough storage, tcpout output queue can persist all events to disk instead of buying expensive third party subscription(unsupported) to persist data to SQS/S3.
-
If there are two tcpout output destinations and one is down for extended period. Down destination has large enough PQ to persist data, then second destination is not blocked. Second destination will block only once PQ of down destination is full.
- Don't have to pay for third party SQS & S3 puts.
- Third party/ external S3 persistent queue introduces permanent additional latency( due to detour to external SQS/S3 queue). There are chances of loss of events( events getting in to SQS/S3 DLQ).
- Third party/ external SQS/S3 persistent queuing requires batching events, which adds additional latency in order to reduce SQS/S3 puts cost.
- Unwanted additional network bandwidth usage incurred due to uploading all data to SQS/S3 and then downloading .
- Third party imposes upload payload size limits.
- Monitored corporate laptops are off network, not connected to internet or not connected to VPN for extended period of time. Later laptops might get switched off but events should be persisted and forwarded as and when laptop connects to network.
- Sensitive data should stay/persisted within network.
- On demand persistent queuing on forwarding tier when Indexer Clustering is down.
- On demand persistent queuing on forwarding tier when Indexer Clustering indexing is slow due to high system load.
- On demand persistent queuing on forwarding tier when Indexer Clustering is in rolling restart.
- On demand persistent queuing on forwarding tier during Indexer Clustering upgrade.
- Don't have to use decade old S2S protocol version as suggested by some third party vendors ( you all know enableOldS2SProtocol=true in outputs.conf)
How to enable?
Just set persistentQueueSize as per outputs.conf
[tcpout:splunk-group1]
persistentQueueSize=1TB
[tcpout:splunk-group2]
persistentQueueSize=2TBNote: Sizing guide. Run following SPL
index=_internal source=*metrics.log* group=tcpin_connections hostname=<all IHF> host=<all idx>| stats sum(kb) as required_pq_in_kb by hostname Run above search for number of days/hours/minutes depending on how much of PQ is needed. required_pq_in_kb is the size of approximate overall PQ on a given IHF. needed. required_pq_in_kb/parallelIngestionPipelines is the approximate value for PQ for each pipeline.
persistentQueueSize = required_pq_in_kb/parallelIngestionPipelines
