File system hangs after ML1 release
Our filesystem was overflowing with files with an .nfsxxx extension since the implementation of the 184.108.40.206 (v6.2 SP2 ML1) release on 10th of August. Before that date issue was never experienced.
The applicaiton places ZIP-files on the local file system, they are then moved to another location by an Activity. The move of the ZIP-files succeeds, but for most of the files, this also provides .nfsxxx files. Eventually (19th of August) the maximum number of opened files was reached on the file system. The Pega Production node was unresponsive and therefore needed to reboot.
No specific error message. The node became unresponsive. File system was full of these .nfsxxx files.
Steps to Reproduce
Normal production work involving file copies from local filesystem to NFS mounted target destination.
Environment Issue. Customer confirmed that moving from NFS v3 to NFS v4 corrected the behaviour observed.
PRPC does not depend on NFS and it does not have code specific for NFS.
The use case in question is completely custom Java and does not use PRPC Rule-Connect-File.
Various articles and forum threads on the web report similar issue scenarios with NFS (specifically NFSv3). The symptoms reported detail an issue where temporary files remain open in the O/S causing the maximum number of files open limit to be reached. Pega GCS have reviewed the details and confirm that PRPC is not a causal factor in these interactions. Upgrading from NFSv3 to NFSv4 appeared to negate the problem behaviour.