Method Details

• getAvgPeerTestTime

  public int getAvgPeerTestTime()

  Average successful peer test time over the last minute.

  Returns:

  average test time in milliseconds, or 0 if no context available

• getTotalAvgPeerTestTime

  public int getTotalAvgPeerTestTime()

  Total average peer test time over the last hour, including slow tests.

  Returns:

  total average test time in milliseconds, or 0 if no context available

• getPeerTestDelay

  private long getPeerTestDelay()

  Calculates the delay before starting the next round of peer tests.

  Three-phase decay based on uptime:

  • 0–4 hours (startup): 3s — aggressively profile all fast peers
  • 4–12 hours (warm-up): 5s — ramp down while keeping data fresh
  • 12+ hours (steady): 8s — maintenance cadence

  User override via PROP_PEER_TEST_DELAY is respected for the steady-state value only; startup and warm-up phases use hardcoded values.

  Returns:

  delay in milliseconds before next test round

• getTestTimeout

  private int getTestTimeout()

  Determines the timeout for individual peer tests with automatic adjustment.

  Ensures timeout is never set below the average successful test time to avoid false timeouts during normal operation.

  Returns:

  timeout in milliseconds, adjusted if necessary

• getTestConcurrency

  private int getTestConcurrency()

  Determines the number of peers to test concurrently based on router uptime.

  Three-phase decay:

  • 0–4 hours (startup): 4 — profile all fast peers quickly
  • 4–12 hours (warm-up): 3 — ramp down
  • 12+ hours (steady): 1 — maintenance cadence

  Concurrency is limited to 1 when CPU load exceeds 95% to prevent system overload.

  Returns:

  number of peers to test in parallel

• startTesting

  public void startTesting(PeerManager manager)

  Starts the peer testing process.

  Schedules the first test run with the configured delay. If uptime is under 3 minutes, the log notes that testing will commence shortly.

  Parameters:

  manager - the peer manager to use for peer selection

• stopTesting

  public void stopTesting()

  Stops the peer testing process gracefully.

  The current test round will complete, but no new rounds will be scheduled.

• schedulePriorityTests

  public void schedulePriorityTests(List<Hash> peers)

  Schedule peers for priority testing (e.g., from slow tunnels). These peers will be tested on the next test cycle before other peers.

  Parameters:

  peers - list of peer hashes to test urgently

• getName

  public String getName()

  Name of this job.

  Returns:

  the name this job is listed under in the job queue and the logs, "Test Peers"

• runJob

  public void runJob()

  Main job execution loop that performs peer testing with adaptive scheduling.

  Process flow:

  1. Check if testing should continue
  2. Select peers for testing based on criteria
  3. Test each selected peer
  4. Adapt next run delay based on system conditions

  Adaptive Behavior:

  • Skip the round and multiply delay by 5 if job lag > 2000ms
  • Scale delay proportionally to lag (1 + lag/500, capped at 5 min) if job lag > 300ms or CPU load > 80%
  • Normal delay otherwise

• selectPeersToTest

  private Set<RouterInfo> selectPeersToTest()

  Selects peers for testing based on performance and capability criteria.

  Selection Criteria:

  • Priority: peers recorded from slow tunnels, tested first
  • Regular: all profiled peers with a RouterInfo and version 0.9.57+, sorted by tier priority then staleness
  • Excluded: Missing local RouterInfo or profile, or a version older than 0.9.57

  Performance Impact:

  • Uses cached lookups to minimize database queries
  • Pre-parses capabilities to avoid repeated string operations
  • Logs skipped peers with specific reasons for debugging

  Returns:

  set of RouterInfo structures for testing (excluding self)

• testPeer

  private void testPeer(RouterInfo peer)

  Fire off the necessary jobs and messages to test the given peer The message is a store of the peer's RI to itself, with a reply token.

  Parameters:

  peer - the peer to test

• getOutboundTunnelId

  private TunnelInfo getOutboundTunnelId()

  What tunnel will we send the test out through?

  Returns:

  tunnel id (or null if none are found)

• getInboundTunnelId

  private TunnelInfo getInboundTunnelId()

  What tunnel will we get replies through?

  Returns:

  tunnel id (or null if none are found)

• buildMessage

  private DatabaseStoreMessage buildMessage(RouterInfo peer, TunnelId replyTunnel, Hash replyGateway, long nonce, long expiration)

  Build a message to test the peer with. The message is a store of the peer's RI to itself, with a reply token.

  Parameters:

  peer - the peer whose own RouterInfo is stored back to it

  replyTunnel - the id of our inbound tunnel the reply must return through

  replyGateway - the hash of that tunnel's gateway, asked to forward the reply

  nonce - the reply token echoed in the response so we can match it to this test

  expiration - the message expiration time in ms since epoch

  Returns:

  the assembled DatabaseStoreMessage carrying the reply token

• tierRank

  private static int tierRank(ProfileOrganizer organizer, Hash peer)

  Selection priority for a test candidate: tier first (fast 0, high-cap 1, other 2), then staleness of the persisted tunnel-test EWMA so never-tested peers go first.

  Precomputed by tierRank(ProfileOrganizer, Hash) rather than derived from ProfileOrganizer.isFast/isHighCapacity inside the comparator: those take the reorganize read lock, and calling them per comparison meant 4 lock acquisitions per comparison across an O(n log n) sort.

  Parameters:

  organizer - the profile organizer holding the fast and high-capacity sets

  peer - the test candidate whose tier is ranked

  Returns:

  0 for fast, 1 for high capacity, 2 for everything else