Will do.
On Wed, Mar 7, 2018 at 9:33 AM, Kostas Kloudas
wrote:
> Why not opening a JIRA and working on adding some debug
> statements that you consider useful?
>
> This could help the next user that faces the same issues ;)
>
> Kostas
>
> On Mar 7, 2018, at 3:29 PM, Vishal Santoshi
> wrote:
>
>
Why not opening a JIRA and working on adding some debug
statements that you consider useful?
This could help the next user that faces the same issues ;)
Kostas
> On Mar 7, 2018, at 3:29 PM, Vishal Santoshi wrote:
>
> Aah, yes we never had a sink state so never came across a case where it was
Aah, yes we never had a sink state so never came across a case where it was
ever exercised. When the range expires, it is a prune rather than a stop
state ( we were expecting it to be a stop state ) which is some what
misleading if we hold stop state to " that invalidates a partial match "
whatev
Hi Vishal,
A stopState is a state that invalidates a partial match, e.g.
a.NotFollowedBy(b).followedBy(c).
If you have an “a” and then you see a “b” then you invalidate the pattern.
A finalState is the one where a match has been found.
Kostas
> On Mar 7, 2018, at 3:20 PM, Vishal Santoshi wro
Absolutely. For one a simple m out of n true conditions where n is defined
by range is a little under optimized as in just using time(m) will not
short circuit the partial patterns till the time range is achieved even if
there is no way m true conditions can be achieved ( we already have had n-m
f
Hi Vishal,
Thanks a lot for sharing your experience and the potential caveats to consider
when
specifying your pattern.
I agree that there is room for improvement when it comes to the state
checkpointed in Flink.
We already have some ideas but still, as you also said, the bulk of the space
co
Hello all, There were recent changes to the flink master that I pulled in
and that *seems* to have solved our issue.
Few points
* CEP is heavy as the NFA transition matrix as state which can be
possibly n^2 ( worst case ) can easily blow up space requirements. The
after match skip strateg
We could not recreate in a controlled setup, but here are a few notes that
we have gathered on a simple "times(n),within(..)"
In case where the Event does not create a Final or Stop state
* As an NFA processes an Event, NFA mutates if there is a true Event. Each
computation is a counter that kee
Thanks a lot Vishal!
We are looking forward to a test case that reproduces the failure.
Kostas
> On Feb 2, 2018, at 4:05 PM, Vishal Santoshi wrote:
>
> This is the pattern. Will create a test case.
> /**
> *
> * @param condition a single condition is applied as a acceptance criteria
> *
This is the pattern. Will create a test case.
/**
*
* @param condition a single condition is applied as a acceptance criteria
* @param params defining the bounds of the pattern.
* @param the element in the stream
* @return compiled pattern alonf with the params.
*/
public static RelaxedCo
Could you provide some example to reproduce the case? Or the Pattern that you
are using? It would help track down the issue.
> On 2 Feb 2018, at 13:35, Vishal Santoshi wrote:
>
> I have pulled in the flink master cep library and the runtime ( the cluster )
> is configured to work against the l
I have flink master CEP library code imported to a 1.4 build.
On Thu, Feb 1, 2018 at 5:33 PM, Vishal Santoshi
wrote:
> A new one
>
> java.lang.OutOfMemoryError: Java heap space
> at java.util.Arrays.copyOf(Arrays.java:3332)
> at
> java.lang.AbstractStringBuilder.ensureCapacityInter
A new one
java.lang.OutOfMemoryError: Java heap space
at java.util.Arrays.copyOf(Arrays.java:3332)
at
java.lang.AbstractStringBuilder.ensureCapacityInternal(AbstractStringBuilder.java:124)
at
java.lang.AbstractStringBuilder.append(AbstractStringBuilder.java:448)
a
It happens when it looks to throw an exception and calls
shardBuffer.toString. b'coz of the check
int id = sharedBuffer.entryId;
Preconditions.checkState(id != -1, "Could not find id for entry: " +
*sharedBuffer*);
On Thu, Feb 1, 2018 at 5:09 PM, Vishal Santoshi
wrote:
> The watermark ha
The watermark has not moved for this pattern to succeed ( or other wise ),
the issue though is that it is pretty early in the pipe ( like within a
minute ). I am replaying from a kafka topic but the keyed operator has
emitted no more than 1500 plus elements to SelectCEPOperator ( very visible
on t
This is a pretty simple pattern, as in I hardly have 1500 elements (
across 600 keys at the max ) put in
and though I have a pretty wide range , as in I am looking at a
relaxed pattern ( like 40 true conditions in 6 hours ),
I get this. I have the EventTime turned on.
java.lang.OutOfMemoryErro
Thanks. Confirmed through tests the above behavior.
On Tue, Jan 23, 2018 at 4:09 AM, Kostas Kloudas wrote:
> Hi Vishal,
>
> Thanks for checking and glad to hear that your job works after the fix!
>
> As for the equals/hashcode question, if your question is if you have to
> implement exact equals
Hi Vishal,
Thanks for checking and glad to hear that your job works after the fix!
As for the equals/hashcode question, if your question is if you have to
implement exact equals() method and the corresponding hashcode()
then the answer is yes. These methods are used when retrieving and cleaning
Have tested against the 1.5 SNAPShot ( I simply pulled the source code
into my distribution and compiled it into my job jar ). Both the test code
and the cluster seems to work ok. Have not tested the "savepoint and
resume" mode but restore from checkpoint works. I brought the JM down and
restarte
Will do.
On Sun, Jan 14, 2018 at 10:23 AM, Fabian Hueske wrote:
> We don't have a schedule for bugfix releases but do them based on need.
> AFAIK, a discussion about a 1.4.1 release has not been started yet.
>
> Would you like to kick that off by sending a mail to the dev mailing list?
>
>
> 201
We don't have a schedule for bugfix releases but do them based on need.
AFAIK, a discussion about a 1.4.1 release has not been started yet.
Would you like to kick that off by sending a mail to the dev mailing list?
2018-01-12 16:41 GMT+01:00 Vishal Santoshi :
> Thanks. We will.
>
>When is
Thanks. We will.
When is 1.4.1 scheduled for release ?
On Fri, Jan 12, 2018 at 3:24 AM, Dawid Wysakowicz <
wysakowicz.da...@gmail.com> wrote:
> Hi Vishal,
> I think it might be due to this bug: https://issues.apache.org/
> jira/browse/FLINK-8226
> It was merged for 1.4.1 and 1.5.0. Could you
Hi Vishal,
I think it might be due to this bug:
https://issues.apache.org/jira/browse/FLINK-8226
It was merged for 1.4.1 and 1.5.0. Could you check with this changes applied?
Would be really helpful. If the error still persists could you file a jira?
Regards
Dawid
> On 11 Jan 2018, at 19:49, Vi
When checkpointing is turned on a simple CEP loop pattern
private Pattern, ?> alertPattern =
Pattern.>begin("start").where(checkStatusOn)
.followedBy("middle").where(checkStatusOn).times(2)
.next("end").where(checkStatusOn).within(Time.minutes(5))
I see failures.
SimpleBinaryEve
Hi Kieran,
which statebackend are you using for your CEP job? Using RocksDB as a state
backend could potentially fix the issue.
What's the number of keys in your stream?
On Tue, Nov 29, 2016 at 3:18 PM, kieran . wrote:
> Hello,
>
> I am currently building a multi-tenant monitoring application
Hello,
I am currently building a multi-tenant monitoring application and exploring the
effectiveness of different Complex Event Processors (CEP) and whether or not
this would be a potential solution for what I want to achieve. I have created a
small test application which utilises Flink and its
26 matches
Mail list logo
