(crawler) Improve deferred task behavior

The previous icon was of an irregular size and shifted the layout in an unaesthetic way.

code/common/service/java/nu/marginalia/service/module/ServiceConfigurationModule.java

@@ -121,6 +121,7 @@ public class ServiceConfigurationModule extends AbstractModule {
 
         while (nets.hasMoreElements()) {
             NetworkInterface netif = nets.nextElement();
+            logger.info("Considering network interface {}:  Up? {},  Loopback? {}", netif.getDisplayName(), netif.isUp(), netif.isLoopback());
             if (!netif.isUp() || netif.isLoopback()) {
                 continue;
             }
@@ -128,6 +129,7 @@ public class ServiceConfigurationModule extends AbstractModule {
             Enumeration<InetAddress> inetAddresses = netif.getInetAddresses();
             while (inetAddresses.hasMoreElements()) {
                 InetAddress addr = inetAddresses.nextElement();
+                logger.info("Considering address {}: SiteLocal? {}, Loopback? {}", addr.getHostAddress(), addr.isSiteLocalAddress(), addr.isLoopbackAddress());
                 if (addr.isSiteLocalAddress() && !addr.isLoopbackAddress()) {
                     return addr.getHostAddress();
                 }

code/common/service/java/nu/marginalia/service/server/MetricsServer.java

@@ -13,7 +13,7 @@ import java.net.InetSocketAddress;
 
 public class MetricsServer {
 
-    private static Logger logger = LoggerFactory.getLogger(MetricsServer.class);
+    private static final Logger logger = LoggerFactory.getLogger(MetricsServer.class);
 
     @Inject
     public MetricsServer(ServiceConfiguration configuration) {
@@ -30,6 +30,8 @@ public class MetricsServer {
 
             context.addServlet(new ServletHolder(new MetricsServlet()), "/metrics");
 
+            logger.info("MetricsServer listening on {}:{}", configuration.bindAddress(), configuration.metricsPort());
+
             server.start();
         }
         catch (Exception|NoSuchMethodError ex) {

code/processes/crawling-process/java/nu/marginalia/crawl/CrawlerMain.java

@@ -248,9 +248,14 @@ public class CrawlerMain extends ProcessMainClass {
             // (this happens when the process is restarted after a crash or a shutdown)
             tasksDone.set(workLog.countFinishedJobs());
 
-            // Create crawl tasks and submit them to the pool for execution
+            // List of deferred tasks used to ensure beneficial scheduling of domains with regard to DomainLocks,
+            // merely shuffling the domains tends to lead to a lot of threads being blocked waiting for a semphore,
+            // this will more aggressively attempt to schedule the jobs to avoid blocking
+            List<CrawlTask> taskList = new ArrayList<>();
+
+            // Create crawl tasks
             for (CrawlSpecRecord crawlSpec : crawlSpecRecords) {
-                if (workLog.isJobFinished(crawlSpec.domain()))
+                if (workLog.isJobFinished(crawlSpec.domain))
                     continue;
 
                 var task = new CrawlTask(
@@ -261,11 +266,22 @@ public class CrawlerMain extends ProcessMainClass {
                         domainStateDb,
                         workLog);
 
-                if (pendingCrawlTasks.putIfAbsent(crawlSpec.domain(), task) == null) {
-                    pool.submitQuietly(task);
+                // Try to run immediately, to avoid unnecessarily keeping the entire work set in RAM
+                if (!trySubmitDeferredTask(task)) {
+                    // Otherwise add to the taskList for deferred execution
+                    taskList.add(task);
                 }
             }
 
+             // Schedule viable tasks for execution until list is empty
+            while (!taskList.isEmpty()) {
+                taskList.removeIf(this::trySubmitDeferredTask);
+
+                // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                // we might have no new viable tasks to run for hours on end
+                TimeUnit.MILLISECONDS.sleep(50);
+            }
+
             logger.info("Shutting down the pool, waiting for tasks to complete...");
 
             pool.shutDown();
@@ -290,6 +306,28 @@ public class CrawlerMain extends ProcessMainClass {
         }
     }
 
+    /** Submit a task for execution if it can be run, returns true if it was submitted
+     * or if it can be discarded */
+    private boolean trySubmitDeferredTask(CrawlTask task) {
+        if (!task.canRun()) {
+            return false;
+        }
+
+        if (pendingCrawlTasks.putIfAbsent(task.domain, task) != null) {
+            return true; // task has already run, duplicate in crawl specs
+        }
+
+        try {
+            // This blocks the caller when the pool is full
+            pool.submitQuietly(task);
+            return true;
+        }
+        catch (RuntimeException ex) {
+            logger.error("Failed to submit task " + task.domain, ex);
+            return false;
+        }
+    }
+
     public void runForSingleDomain(String targetDomainName, FileStorageId fileStorageId) throws Exception {
         runForSingleDomain(targetDomainName, fileStorageService.getStorage(fileStorageId).asPath());
     }
@@ -346,9 +384,20 @@ public class CrawlerMain extends ProcessMainClass {
             this.id = Integer.toHexString(domain.hashCode());
         }
 
+        /** Best effort indicator whether we could start this now without getting stuck in
+         * DomainLocks purgatory */
+        public boolean canRun() {
+            return domainLocks.canLock(new EdgeDomain(domain));
+        }
+
         @Override
         public void run() throws Exception {
 
+            if (workLog.isJobFinished(domain)) { // No-Op
+                logger.info("Omitting task {}, as it is already run", domain);
+                return;
+            }
+
             Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
             Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
             Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
@@ -403,7 +452,7 @@ public class CrawlerMain extends ProcessMainClass {
                 logger.error("Error fetching domain " + domain, e);
             }
             finally {
-                // We don't need to double-count these; it's also kept int he workLog
+                // We don't need to double-count these; it's also kept in the workLog
                 pendingCrawlTasks.remove(domain);
                 Thread.currentThread().setName("[idle]");
 
@@ -494,7 +543,7 @@ public class CrawlerMain extends ProcessMainClass {
     //
     // This must be synchronized as chewing through parquet files in parallel leads to enormous memory overhead
     private synchronized Path migrateParquetData(Path inputPath, String domain, Path crawlDataRoot) throws IOException {
-        if (!inputPath.endsWith(".parquet")) {
+        if (!inputPath.toString().endsWith(".parquet")) {
             return inputPath;
         }
 

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/HttpFetcherImpl.java

@@ -251,6 +251,7 @@ public class HttpFetcherImpl implements HttpFetcher {
         return new SitemapRetriever();
     }
 
+    /** Recursively fetch sitemaps */
     @Override
     public List<EdgeUrl> fetchSitemapUrls(String root, CrawlDelayTimer delayTimer) {
         try {
@@ -270,7 +271,7 @@ public class HttpFetcherImpl implements HttpFetcher {
             while (!sitemapQueue.isEmpty() && ret.size() < 20_000 && ++fetchedSitemaps < 10) {
                 var head = sitemapQueue.removeFirst();
 
-                switch (fetchSitemap(head)) {
+                switch (fetchSingleSitemap(head)) {
                     case SitemapResult.SitemapUrls(List<String> urls) -> {
 
                         for (var url : urls) {
@@ -306,7 +307,7 @@ public class HttpFetcherImpl implements HttpFetcher {
     }
 
 
-    private SitemapResult fetchSitemap(EdgeUrl sitemapUrl) throws URISyntaxException, IOException, InterruptedException {
+    private SitemapResult fetchSingleSitemap(EdgeUrl sitemapUrl) throws URISyntaxException, IOException, InterruptedException {
         HttpRequest getRequest = HttpRequest.newBuilder()
                 .GET()
                 .uri(sitemapUrl.asURI())

code/processes/crawling-process/java/nu/marginalia/crawl/logic/DomainLocks.java

@@ -44,6 +44,14 @@ public class DomainLocks {
         return new Semaphore(2);
     }
 
+    public boolean canLock(EdgeDomain domain) {
+        Semaphore sem = locks.get(domain.topDomain.toLowerCase());
+        if (null == sem)
+            return true;
+        else
+            return sem.availablePermits() > 0;
+    }
+
     public static class DomainLock implements AutoCloseable {
         private final String domainName;
         private final Semaphore semaphore;

code/processes/crawling-process/model/java/nu/marginalia/io/SerializableCrawlDataStream.java

@@ -42,18 +42,20 @@ public interface SerializableCrawlDataStream extends AutoCloseable {
     {
 
         String fileName = fullPath.getFileName().toString();
-        if (fileName.endsWith(".parquet")) {
+
+        if (fileName.endsWith(".slop.zip")) {
             try {
-                return new ParquetSerializableCrawlDataStream(fullPath);
+                return new SlopSerializableCrawlDataStream(fullPath);
             } catch (Exception ex) {
                 logger.error("Error reading domain data from " + fullPath, ex);
                 return SerializableCrawlDataStream.empty();
             }
         }
 
-        if (fileName.endsWith(".slop.zip")) {
+        else if (fileName.endsWith(".parquet")) {
+            logger.error("Opening deprecated parquet-style crawl data stream", new Exception());
             try {
-                return new SlopSerializableCrawlDataStream(fullPath);
+                return new ParquetSerializableCrawlDataStream(fullPath);
             } catch (Exception ex) {
                 logger.error("Error reading domain data from " + fullPath, ex);
                 return SerializableCrawlDataStream.empty();

code/services-application/search-service/java/nu/marginalia/search/model/SearchFilters.java

@@ -81,6 +81,7 @@ public class SearchFilters {
                             ),
                             List.of(
                                     new Filter("Vintage", "fa-clock-rotate-left", SearchProfile.VINTAGE, parameters),
+                                    new Filter("Small Web", "fa-minus", SearchProfile.SMALLWEB, parameters),
                                     new Filter("Plain Text", "fa-file", SearchProfile.PLAIN_TEXT, parameters),
                                     new Filter("Tilde", "fa-house", SearchProfile.TILDE, parameters)
                             ),

code/services-application/search-service/resources/jte/part/testenvHead.jte

@@ -9,6 +9,14 @@
                     nicotine: '#f8f8ee',
                     margeblue: '#3e5f6f',
                     liteblue: '#0066cc',
+                },
+                screens: {
+                    'coarsepointer': {
+                        'raw': '(pointer: coarse)'
+                    },
+                    'finepointer': {
+                        'raw': '(pointer: fine)'
+                    },
                 }
             },
             screens: {

code/services-application/search-service/resources/jte/serp/main.jte

@@ -23,7 +23,7 @@
                 @template.serp.part.searchform(query = results.getParams().query(), profile = results.getProfile(), filters = results.getFilters())
             </div>
             <div class="grow"></div>
-            <button class="fixed bottom-10 right-5 sm:hidden text-sm bg-margeblue text-white p-4 rounded-xl active:text-slate-200" id="filter-button">
+            <button class="fixed bottom-10 right-5 finepointer:hidden md:hidden text-sm bg-margeblue text-white p-4 rounded-xl active:text-slate-200" id="filter-button">
                 <i class="fas fa-filter mr-3"></i>
                 Filters
             </button>

code/services-application/search-service/resources/jte/serp/part/sidebar.jte

@@ -3,7 +3,7 @@
 
 @param SearchFilters filters
 
-<aside class="md:w-64 py-4 shrink-0 hidden sm:block">
+<aside class="md:w-64 py-4 shrink-0 hidden md:block finepointer:block">
     <div class="space-y-6 sticky top-4">
         <div class="bg-white dark:bg-gray-800 p-4 border dark:border-gray-600 border-gray-300">
             <h2 class="font-medium mb-3 flex items-center font-serif hidden md:block">

code/services-application/search-service/tailwind/tailwind.config.js

@@ -9,6 +9,14 @@ module.exports = {
         nicotine: '#f8f8ee',
         margeblue: '#3e5f6f',
         liteblue: '#0066cc',
+      },
+      screens: {
+        'coarsepointer': {
+          'raw': '(pointer: coarse)'
+        },
+        'finepointer': {
+          'raw': '(pointer: fine)'
+        },
       }
     },
     screens: {