Author: jkreps
Date: Thu Mar 6 20:22:04 2014
New Revision: 1575028
URL: http://svn.apache.org/r1575028
Log:
Fix typo and clarify log compaction guarantees.
Modified:
kafka/site/081/design.html
Modified: kafka/site/081/design.html
URL: http://svn.apache.org/viewvc/kafka/site/081/design.html?rev=1575028&r1=1575027&r2=1575028&view=diff
==============================================================================
--- kafka/site/081/design.html (original)
+++ kafka/site/081/design.html Thu Mar 6 20:22:04 2014
@@ -274,9 +274,8 @@ Log compaction guarantees the following:
<li>Any consumer that stays caught-up to within the head of the log will see a message
for each offset.
<li>Message order is always maintained (that is, compaction will never re-order messages
just remove some).
<li>The offset for a message never changes, it is the permanent identifier for a position
in the log.
-<li>Any read from offset 0 will contain see the final state of all records provided
it completes in less than a configurable time period (even if further compaction occurs during
this read).
+<li>Any read from offset 0 will see the final state of all records (in order) provided
it reaches the head of the log in a time period less than the topic's delete.retention.ms
setting (the default is 24 hours). This is important as delete marker removal happens concurrently
with read.
</ol>
-The third guarantee seems somewhat esoteric but this ensures that a reader will end in a
coherent state (equivalent to reading the full uncompacted log) even though the underlying
log may be compacted while this read occurs.
<h4>Log Compaction Details</h4>
|