Thank you for your question. SnapLogic pools SFTP sessions at the node (JVM) level. It does not create a pool per pipeline execution, and each SSH/SFTP-enabled Snap does not manage its own separate pool.
Scope
• Session pooling applies primarily to Binary Snaps (for example File Reader, File Writer and Directory Browser) when they use the SFTP protocol. Hadoop Snaps may also use it.
• The pool is per node, not per Snaplex. Pipelines and Snaps running on the same node share it.
How it works
1. A Snap obtains a session to the SFTP server, then opens a channel on that session.
2. Pooling is enabled by default. Sessions are keyed on server credentials: host, port, username and password. Snaps that connect with the same credentials can reuse a session, even if they use different SnapLogic accounts.
3. If an idle session is available, it is used. A session found to be dead is discarded and the next idle session is tried. If none are active, a new session is created. After use, the session returns to the cache.
4. By default each session carries one channel, so one session cannot serve multiple Snaps at the same time. Concurrent Snaps with the same credentials therefore open separate sessions unless the channel limit is raised.
5. A session idle past its timeout is discarded.
Configuration (JVM options on the node)
• cacheSftpSessions: set to false to disable pooling (default: enabled).
• maxSftpChannels: maximum channels per session. The default is 1. Raise it to let one session be shared by multiple Snaps at once.
• sftpSessionTimeoutSec: idle timeout after last use (default 3600 seconds, i.e. 1 hour).
