From ebdb28ee4bd6c11c4947fbe49c1ddd9fe02652b6 Mon Sep 17 00:00:00 2001
From: Tyler Ouyang <tylerwowen@users.noreply.github.com>
Date: Tue, 1 Aug 2023 09:57:09 -0700
Subject: [PATCH] Fix AgentJanitor bugs (#1225)

I previously mis-used Java stream. Fixed them and added more logging.

Validated in pre-prod
---
 .../pinterest/teletraan/worker/AgentJanitor.java  | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/deploy-service/teletraanservice/src/main/java/com/pinterest/teletraan/worker/AgentJanitor.java b/deploy-service/teletraanservice/src/main/java/com/pinterest/teletraan/worker/AgentJanitor.java
index 011e50ac8c..df830edbc9 100644
--- a/deploy-service/teletraanservice/src/main/java/com/pinterest/teletraan/worker/AgentJanitor.java
+++ b/deploy-service/teletraanservice/src/main/java/com/pinterest/teletraan/worker/AgentJanitor.java
@@ -23,6 +23,7 @@
 import java.util.Map;
 import java.util.Set;
 import java.util.concurrent.TimeUnit;
+import java.util.stream.Collectors;
 
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -122,16 +123,18 @@ private boolean isHostStale(HostAgentBean hostAgentBean) {
      */
     private void determineStaleHostCandidates() {
         long minThreshold = janitorStartTime - minStaleHostThreshold;
-        long maxThreshold = janitorStartTime - maxLaunchLatencyThreshold;
+        long maxThreshold = janitorStartTime - maxStaleHostThreshold;
         List<HostAgentBean> unreachableHosts;
         try {
+            LOG.debug("getting hosts between {}, {}", maxThreshold, minThreshold);
             unreachableHosts = hostAgentDAO.getStaleHosts(maxThreshold, minThreshold);
         } catch (Exception ex) {
             LOG.error("failed to get unreachable hosts", ex);
             return;
         }
-        ArrayList<String> unreachableHostIds = new ArrayList<>();
-        unreachableHosts.stream().map(hostAgent -> unreachableHostIds.add(hostAgent.getHost_id()));
+        List<String> unreachableHostIds = unreachableHosts.stream().map(HostAgentBean::getHost_id)
+                .collect(Collectors.toList());
+        LOG.debug("fetched {} unreachable hosts", unreachableHostIds.size());
 
         Set<String> terminatedHosts = getTerminatedHostsFromSource(unreachableHostIds);
         for (String unreachableId : unreachableHostIds) {
@@ -152,6 +155,7 @@ private void processStaleHosts() {
         long maxThreshold = janitorStartTime - maxStaleHostThreshold;
         List<HostAgentBean> staleHosts;
         try {
+            LOG.debug("getting hosts before {}", maxThreshold);
             staleHosts = hostAgentDAO.getStaleHosts(maxThreshold);
         } catch (Exception ex) {
             LOG.error("failed to get stale hosts", ex);
@@ -159,7 +163,8 @@ private void processStaleHosts() {
         }
 
         Map<String, HostAgentBean> staleHostMap = new HashMap<>();
-        staleHosts.stream().map(hostAgent -> staleHostMap.put(hostAgent.getHost_id(), hostAgent));
+        staleHosts.stream().forEach(hostAgent -> staleHostMap.put(hostAgent.getHost_id(), hostAgent));
+        LOG.debug("fetched {} unreachable hosts", staleHostMap.values().size());
 
         Set<String> terminatedHosts = getTerminatedHostsFromSource(new ArrayList<>(staleHostMap.keySet()));
         for (String staleId : staleHostMap.keySet()) {
@@ -170,6 +175,8 @@ private void processStaleHosts() {
                 if (isHostStale(hostAgent)) {
                     LOG.warn("Agent ({}) is stale (not Pinging Teletraan), but might be running.",
                             hostAgent);
+                } else {
+                    LOG.debug("host {} is not stale", staleId);
                 }
             }
         }