zacharymorn commented on a change in pull request #128:
URL: https://github.com/apache/lucene/pull/128#discussion_r642646093
##########
File path: lucene/core/src/java/org/apache/lucene/index/CheckIndex.java
##########
@@ -605,209 +680,103 @@ public Status checkIndex(List<String> onlySegments)
throws IOException {
result.newSegments.clear();
result.maxSegmentName = -1;
- for (int i = 0; i < numSegments; i++) {
- final SegmentCommitInfo info = sis.info(i);
- long segmentName = Long.parseLong(info.info.name.substring(1),
Character.MAX_RADIX);
- if (segmentName > result.maxSegmentName) {
- result.maxSegmentName = segmentName;
- }
- if (onlySegments != null && !onlySegments.contains(info.info.name)) {
- continue;
- }
- Status.SegmentInfoStatus segInfoStat = new Status.SegmentInfoStatus();
- result.segmentInfos.add(segInfoStat);
- msg(
- infoStream,
- " "
- + (1 + i)
- + " of "
- + numSegments
- + ": name="
- + info.info.name
- + " maxDoc="
- + info.info.maxDoc());
- segInfoStat.name = info.info.name;
- segInfoStat.maxDoc = info.info.maxDoc();
-
- final Version version = info.info.getVersion();
- if (info.info.maxDoc() <= 0) {
- throw new RuntimeException("illegal number of documents: maxDoc=" +
info.info.maxDoc());
- }
-
- int toLoseDocCount = info.info.maxDoc();
-
- SegmentReader reader = null;
-
- try {
- msg(infoStream, " version=" + (version == null ? "3.0" : version));
- msg(infoStream, " id=" +
StringHelper.idToString(info.info.getId()));
- final Codec codec = info.info.getCodec();
- msg(infoStream, " codec=" + codec);
- segInfoStat.codec = codec;
- msg(infoStream, " compound=" + info.info.getUseCompoundFile());
- segInfoStat.compound = info.info.getUseCompoundFile();
- msg(infoStream, " numFiles=" + info.files().size());
- Sort indexSort = info.info.getIndexSort();
- if (indexSort != null) {
- msg(infoStream, " sort=" + indexSort);
- }
- segInfoStat.numFiles = info.files().size();
- segInfoStat.sizeMB = info.sizeInBytes() / (1024. * 1024.);
- msg(infoStream, " size (MB)=" + nf.format(segInfoStat.sizeMB));
- Map<String, String> diagnostics = info.info.getDiagnostics();
- segInfoStat.diagnostics = diagnostics;
- if (diagnostics.size() > 0) {
- msg(infoStream, " diagnostics = " + diagnostics);
+ // checks segments sequentially
+ if (executorService == null) {
+ for (int i = 0; i < numSegments; i++) {
+ final SegmentCommitInfo info = sis.info(i);
+ updateMaxSegmentName(result, info);
+ if (onlySegments != null && !onlySegments.contains(info.info.name)) {
+ continue;
}
- if (!info.hasDeletions()) {
- msg(infoStream, " no deletions");
- segInfoStat.hasDeletions = false;
- } else {
- msg(infoStream, " has deletions [delGen=" + info.getDelGen() +
"]");
- segInfoStat.hasDeletions = true;
- segInfoStat.deletionsGen = info.getDelGen();
- }
-
- long startOpenReaderNS = System.nanoTime();
- if (infoStream != null) infoStream.print(" test: open
reader.........");
- reader = new SegmentReader(info, sis.getIndexCreatedVersionMajor(),
IOContext.DEFAULT);
msg(
infoStream,
- String.format(
- Locale.ROOT, "OK [took %.3f sec]", nsToSec(System.nanoTime() -
startOpenReaderNS)));
+ (1 + i)
+ + " of "
+ + numSegments
+ + ": name="
+ + info.info.name
+ + " maxDoc="
+ + info.info.maxDoc());
+ Status.SegmentInfoStatus segmentInfoStatus = testSegment(sis, info,
infoStream);
+
+ processSegmentInfoStatusResult(result, info, segmentInfoStatus);
+ }
+ } else {
+ ByteArrayOutputStream[] outputs = new ByteArrayOutputStream[numSegments];
+ @SuppressWarnings({"unchecked", "rawtypes"})
+ CompletableFuture<Status.SegmentInfoStatus>[] futures = new
CompletableFuture[numSegments];
+
+ // checks segments concurrently
+ for (int i = 0; i < numSegments; i++) {
+ final SegmentCommitInfo info = sis.info(i);
+ updateMaxSegmentName(result, info);
+ if (onlySegments != null && !onlySegments.contains(info.info.name)) {
+ continue;
+ }
- segInfoStat.openReaderPassed = true;
+ SegmentInfos finalSis = sis;
- long startIntegrityNS = System.nanoTime();
- if (infoStream != null) infoStream.print(" test: check
integrity.....");
- reader.checkIntegrity();
+ ByteArrayOutputStream output = new ByteArrayOutputStream();
+ PrintStream stream;
+ if (i > 0) {
+ // buffer the messages for segment starting from the 2nd one so that
they can later be
+ // printed in order
+ stream = new PrintStream(output, true, IOUtils.UTF_8);
+ } else {
+ // optimize for first segment to print real-time
Review comment:
> Hmm, could we instead just output the full segment's output as each
finishes? So we would presumably see small segments finish first, then all
checks for that small segment are printed?
> Also, note that the "first segment" might not necessarily always be the
largest! The way our default merge policy works, that tends to be the case, but
users can customize the merge policy, and we recently added a feature to let
IndexWriter customize how it sorts the segments. We should maybe sort the
segments by descending docCount and work on them (concurrently) in that order?
Hmm ok, and good to know about the custom segment sorting feature! The
current implementation actually ensures the output to follow the same segment
order as if it were executed sequentially, hence the little trick there to
print first segment output early. But yeah I guess this order requirement is
not really needed, especially under concurrent setting. Let me update that to
give better output (and maybe even fail faster if smaller segment encounters
error early!).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
