@btshivanand
I keep the flowcharts published at https://wiki.splunk.com/Community:HowIndexingWorks handy as a reference.
Look at:
index=_internal source=*metrics.log* host=<yourhost> blocked=true
parsingQueue itself could be full, but as @isoutamo noted, the most likely cause is a blocked output queue, which in turn blocks indexQueue, typingQueue, aggQueue, and parsingQueue on the forwarder.
If an output queue is blocked, check the receiver using the same process. If indexQueue is blocked on an indexer, make sure 1) you have enough free disk space for hot data and 2) your storage is fast enough to keep up with your aggregate ingest rate amortized across your indexer or cluster.
If your heavy forwarder has multiple outputs, a single blocked output will eventually block indexQueue, which will in turn block all outputs in the pipeline. This is especially common with syslog outputs, which have a fixed queue length / buffer size.
