1
1
mirror of https://github.com/MarginaliaSearch/MarginaliaSearch.git synced 2025-10-06 07:32:38 +02:00

Compare commits

...

6 Commits

Author SHA1 Message Date
Viktor Lofgren
62ba30bacf (common) Log info about metrics server 2025-03-10 13:12:39 +01:00
Viktor Lofgren
3bb84eb206 (common) Log info about metrics server 2025-03-10 13:03:48 +01:00
Viktor Lofgren
be7d13ccce (crawler) Correct task execution logic in crawler
The old behavior would flag domains as pending too soon, leading to them being omitted from execution if they were not immediately available to run.
2025-03-09 13:47:51 +01:00
Viktor Lofgren
8c088a7c0b (crawler) Remove custom thread factory
This was causing issues, and not really doing much of benefit.
2025-03-09 11:50:52 +01:00
Viktor Lofgren
ea9a642b9b (crawler) More effective task scheduling in the crawler
This should hopefully allow more threads to be busy
2025-03-09 11:44:59 +01:00
Viktor Lofgren
27f528af6a (search) Fix "Remove Javascript" toggle
A bug was introduced at some point where the special keyword for filtering on javascript was changed to special:scripts, from js:true/js:false.

Solves issue #155
2025-02-28 12:03:04 +01:00
7 changed files with 55 additions and 19 deletions

View File

@@ -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();
}

View File

@@ -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) {

View File

@@ -41,10 +41,7 @@ import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.StandardCopyOption;
import java.security.Security;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.Map;
import java.util.*;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicInteger;
@@ -248,22 +245,47 @@ public class CrawlerMain extends ProcessMainClass {
// (this happens when the process is restarted after a crash or a shutdown)
tasksDone.set(workLog.countFinishedJobs());
// 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> deferredTasks = new LinkedList<>();
// Create crawl tasks and submit them to the pool for execution
for (CrawlSpecRecord crawlSpec : crawlSpecRecords) {
if (workLog.isJobFinished(crawlSpec.domain()))
continue;
var task = new CrawlTask(
// Add to the end of the deferral list
deferredTasks.addLast(new CrawlTask(
crawlSpec,
anchorTagsSource,
outputDir,
warcArchiver,
domainStateDb,
workLog);
workLog));
if (pendingCrawlTasks.putIfAbsent(crawlSpec.domain(), task) == null) {
pool.submitQuietly(task);
}
// Start every task we currently can from the deferral list
deferredTasks.removeIf(task -> {
if (task.canRun()) {
if (pendingCrawlTasks.putIfAbsent(crawlSpec.domain(), task) != null) {
return true; // task has already run, duplicate in crawl specs
}
// This blocks the caller when the pool is full
pool.submitQuietly(task);
return true;
}
return false;
});
}
// Schedule any lingering tasks for immediate execution
for (var task : deferredTasks) {
if (pendingCrawlTasks.putIfAbsent(task.domain, task) != null)
continue;
pool.submitQuietly(task);
}
logger.info("Shutting down the pool, waiting for tasks to complete...");
@@ -346,6 +368,12 @@ 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 {

View File

@@ -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())

View File

@@ -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;

View File

@@ -7,8 +7,7 @@ import java.util.Arrays;
public enum SearchJsParameter {
DEFAULT("default"),
DENY_JS("no-js", "js:true"),
REQUIRE_JS("yes-js", "js:false");
DENY_JS("no-js", "special:scripts");
public final String value;
public final String[] implictExcludeSearchTerms;
@@ -20,7 +19,6 @@ public enum SearchJsParameter {
public static SearchJsParameter parse(@Nullable String value) {
if (DENY_JS.value.equals(value)) return DENY_JS;
if (REQUIRE_JS.value.equals(value)) return REQUIRE_JS;
return DEFAULT;
}

View File

@@ -7,9 +7,7 @@ import java.util.Arrays;
public enum SearchJsParameter {
DEFAULT("default"),
DENY_JS("no-js", "js:true"),
REQUIRE_JS("yes-js", "js:false");
DENY_JS("no-js", "special:scripts");
public final String value;
public final String[] implictExcludeSearchTerms;
@@ -20,7 +18,6 @@ public enum SearchJsParameter {
public static SearchJsParameter parse(@Nullable String value) {
if (DENY_JS.value.equals(value)) return DENY_JS;
if (REQUIRE_JS.value.equals(value)) return REQUIRE_JS;
return DEFAULT;
}