Fix concurrent modification on logs and tags#500
Fix concurrent modification on logs and tags#500pavolloffay merged 3 commits intojaegertracing:masterfrom
Conversation
ghost
assigned
| JaegerSpan span = tracer.buildSpan("foo").start(); | ||
| span.log("foo"); | ||
| for (LogData log: span.getLogs()) { | ||
| span.log("foo2"); |
There was a problem hiding this comment.
Is this test failing when ran against the current master?
Codecov Report
@@ Coverage Diff @@
## master #500 +/- ##
===========================================
- Coverage 88.21% 88.2% -0.02%
Complexity 498 498
===========================================
Files 65 65
Lines 1859 1848 -11
Branches 242 239 -3
===========================================
- Hits 1640 1630 -10
Misses 141 141
+ Partials 78 77 -1
Continue to review full report at Codecov.
|
| this.startTimeNanoTicks = startTimeNanoTicks; | ||
| this.computeDurationViaNanoTicks = computeDurationViaNanoTicks; | ||
| this.tags = new HashMap<String, Object>(); | ||
| this.tags = new ConcurrentHashMap<String, Object>(); |
There was a problem hiding this comment.
I think we could also remove the synchronized from the tags getter.
There was a problem hiding this comment.
I think it could be removed from other methods too. I will rather do it in a separate PR with more concurrent test and keep this clear. synchronized was not related to the concurrent modifications in this case.
|
I am not clear why copy on write is better than normal synchronization. |
|
@pavolloffay I've just commented on the issue #334 with an update on my repo. |
|
thanks @bygui86, I will fix this anyway as it's a bug |
|
Thanks a lot @pavolloffay :) |
pavolloffay
force-pushed
the
fix-concurrent-exception
branch
from
7e094ca to
2dc6656
Compare
Signed-off-by: Pavol Loffay <ploffay@redhat.com>
pavolloffay
force-pushed
the
fix-concurrent-exception
branch
from
2dc6656 to
0005f5e
Compare
|
@yurishkuro @vprithvi I have changed PR to use only wrap the collection in getters. In the next PR I will try to get rid of synchronizing on this with a test validating better performance. |
Signed-off-by: Pavol Loffay <ploffay@redhat.com>
yurishkuro
left a comment
yurishkuro
left a comment
There was a problem hiding this comment.
The changes in getters make sense, what was the change to log methods about? Just refactoring?
| private final Map<String, ?> fields; | ||
|
|
||
| LogData(long time, String message) { | ||
| LogData(long time, Map<String, ?> fields, String message) { |
There was a problem hiding this comment.
nit: I would have preferred the message parameter to come before fields
| return log(timestampMicroseconds, null, event); | ||
| } | ||
|
|
||
| private JaegerSpan log(long timestampMicroseconds, Map<String, ?> fields, String event) { |
There was a problem hiding this comment.
nit: I would have preferred the message parameter to come before fields
|
Just refactoring, I think the current version is cleaner - only one method for processing logs. I will update PR to change the parameter order and merge if nobody objects |
Signed-off-by: Pavol Loffay <ploffay@redhat.com>
5 participants
Fixes #334
Signed-off-by: Pavol Loffay ploffay@redhat.com