kishorekasi commented on a change in pull request #1233: Reduce running time
for testLedgerCreateAdvWithLedgerIdInLoop
URL: https://github.com/apache/bookkeeper/pull/1233#discussion_r173033139
##########
File path:
bookkeeper-server/src/test/java/org/apache/bookkeeper/client/BookieWriteLedgerTest.java
##########
@@ -579,45 +583,64 @@ public void addComplete(int rc, LedgerHandle lh, long
entryId, Object ctx) {
*/
@Test
public void testLedgerCreateAdvWithLedgerIdInLoop() throws Exception {
- long ledgerId;
int ledgerCount = 40;
- List<List<byte[]>> entryList = new ArrayList<List<byte[]>>();
- LedgerHandle[] lhArray = new LedgerHandle[ledgerCount];
-
- List<byte[]> tmpEntry;
- for (int lc = 0; lc < ledgerCount; lc++) {
- tmpEntry = new ArrayList<byte[]>();
-
- ledgerId = rng.nextLong();
- ledgerId &= Long.MAX_VALUE;
- if
(!baseConf.getLedgerManagerFactoryClass().equals(LongHierarchicalLedgerManagerFactory.class))
{
- // since LongHierarchicalLedgerManager supports ledgerIds of
decimal length upto 19 digits but other
- // LedgerManagers only upto 10 decimals
- ledgerId %= 9999999999L;
- }
-
- LOG.info("Iteration: {} LedgerId: {}", lc, ledgerId);
- lh = bkc.createLedgerAdv(ledgerId, 5, 3, 2, digestType,
ledgerPassword, null);
- lhArray[lc] = lh;
-
- for (int i = 0; i < numEntriesToWrite; i++) {
- ByteBuffer entry = ByteBuffer.allocate(4);
- entry.putInt(rng.nextInt(maxInt));
- entry.position(0);
- tmpEntry.add(entry.array());
- lh.addEntry(i, entry.array());
- }
- entryList.add(tmpEntry);
- }
- for (int lc = 0; lc < ledgerCount; lc++) {
- // Read and verify
- long lid = lhArray[lc].getId();
- LOG.info("readEntries for lc: {} ledgerId: {} ", lc,
lhArray[lc].getId());
- readEntries(lhArray[lc], entryList.get(lc));
- lhArray[lc].close();
- bkc.deleteLedger(lid);
+ long maxId = 9999999999L;
+ if
(baseConf.getLedgerManagerFactoryClass().equals(LongHierarchicalLedgerManagerFactory.class))
{
+ // since LongHierarchicalLedgerManager supports ledgerIds of
decimal length upto 19 digits but other
+ // LedgerManagers only upto 10 decimals
+ maxId = Long.MAX_VALUE;
}
+
+ rng.longs(ledgerCount, 0, maxId) // generate a stream of ledger ids
+ .mapToObj(ledgerId -> { // create a ledger for each ledger id
+ LOG.info("Creating adv ledger with id {}", ledgerId);
+ return bkc.newCreateLedgerOp()
+
.withEnsembleSize(1).withWriteQuorumSize(1).withAckQuorumSize(1)
+
.withDigestType(org.apache.bookkeeper.client.api.DigestType.CRC32)
+
.withPassword(ledgerPassword).makeAdv().withLedgerId(ledgerId)
+ .execute()
+ .thenApply(writer -> { // Add entries to ledger when
created
+ LOG.info("Writing stream of {} entries to {}",
+ numEntriesToWrite, ledgerId);
+ List<ByteBuf> entries =
rng.ints(numEntriesToWrite, 0, maxInt)
+ .mapToObj(i -> {
+ ByteBuf entry = Unpooled.buffer(4);
+ entry.retain();
Review comment:
Let me start with saying - "Yes" ByteBuf and Refcounting usage should add
little more context in the comments as to when and where a refcounted object is
release and similarly comments in APIs releasing such objects.
Coming to the good practice of using refcounted objects. For easy coding and
debugging -
Do's
- If the ref counted object is cached (stashed away) for later use retain()
a refcount
- The instance/class/context that retain()s has to release() the refcount
Don't
- retain() refcount in instance/class/context A and release() in B
(exceptions should have a strong justification)
- cache object reference without retain()/release()
- using retain()/release() in an API which does not cache the object
reference
Without a good (and agreed) programming practice, it will be hard to track
refcount issues leading to either accessing garbage or exceptions. This can
also cause memory oversubscription problems as we have seen in the past.
Note additional rules when using ByteBuf with a ChannelPipeline -
http://netty.io/wiki/new-and-noteworthy-in-4.0.html#bytebuf-is-always-reference-counted
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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
With regards,
Apache Git Services
