We need to switch between hosts in case of unavailability. To do this, we recreate OmmConsumer with new settings (only one OmmConsumer exists at a time). If all hosts are unavailable, the memory leaks out exponentially during 30 attempts, which leads to the process crash. It is empirically established that the memory leak occurs as a result of repeated calls to the OmmConsumer constructor. Advise how to avoid this problem?
25
●2 ●5 ●6
EMA OmmConsumer memory leak in C++ API?
<AHS>
This question requires a premium support ticket to deep investigation. However, every clients in this post are not premium user (TRDC). So they cannot raise a ticket.
I will redirect them to GitHub and extend triage.
25.3k
●87 ●12 ●25
I am not aware of memory leak as you describe - please raise a Premium Support Ticket if you wish to investigate in more detail offline.
Before you do that, I just wanted to check if you aware of the ChannelSet functionality of EMA?
The use of ChannelSets allows you to specify a comma-separated list of channels names. Each Channel definition can point to a different host. Channels in the set will be tried by EMA with each reconnection attempt until a successful connection is made.
Please refer to section 3.1.4 of the EMA C++ Config Guide as well as this article on EMA Config for further details.
A similar use case results in a memory leak in the Java API
This would be the case if you are not a named user for your organisation or if your organisation does not have a RDC support membership.
You can check with your Market Data team or your Refinitiv account manager to find out if you do have a named member who can raise a Premium support ticket on your behalf.
In terms of the other Java post, it appears the user opened a Support ticket and closed the ticket after receiving guidance from the RDC team.
If your organisation does not have a Support membership, then you can report your issue on Github and it will be looked at as a best endeavours basis by the support team.
Also, did you try to use the ChannelSets as mentioned above - as this would is the recommended way i.e. letting EMA deal with connectivity issues and failover scenarios - rather than doing it at an application level.
I'm trying. And as far as I understand - the constructor of the consumer in any case will fail after the specified timeout. Which brings us back to the problem of recreating the consumer with further memory leaks and crash. Or if we set an infinite timeout - we have potential problems with stopping the service when all hosts are unavailable (since the constructor of the consumer will never end)
<AHS>
This case has been raised as case 08003411. The Development team is working on this issue.
18
●1 ●2 ●2
Hi Umer,
When I'm following your Premium Support Ticket link I'm getting
Something Went Wrong We were unable to retrieve the page you were trying to reach. We don't want to leave you hanging though. Go back or see other APIs you may be interested in...
Thanks.
I understand that you may not have an RDC account and that the account / admin team is dealing with this?
We don't have RDC account. We have an agreement that our timeOut issue will be looked at as a good will gesture without RDC account. Thanks.
Hello @sergey_at_ark
Please be informed that you can submit the issue to development directly via GitHub issue page.
1
●0 ●0 ●0
I do experience the memory leak as well.
Due to some reasons (not important why) I need to dispose (uninitialize) the OmmConsumer and create a new instance of OmmConsumer. However after roughly 27 iterations (destroy OmmConsumer and create new one) the process consumes 16 GB of memory and crashes. Worth mentioning that this happens even if no clients are registered.
The ChannelSet feature is not what I am looking for and it won't help me.
Thanks for sharing your experience. Please raise a Premium Support Ticket to investigate in more detail offline / raise a bug report.
Thanks.
Thank you for the suggestion however the premium support link you provided does not work for me. Could you please guide me how to reach the form via the webpage's GUI?
Hello @Anton Marček
Please be informed that you can submit the issue to development directly via GitHub issue page too.
7.6k
●15 ●6 ●9
This issue has been reported on GitHub
https://github.com/Refinitiv/Elektron-SDK/issues/118
JIRA ID: ESDK-3495
Based on the latest post from the issue, the development team is working on this issue, and it will be fixed in the next release.
You can change EMA default behavior to log the EMA messages to Stdout as the workaround. We can put EmaConfig.xml to a running directory and then change LoggerType in configuration file to Stdout instead.
DefaultConsumer in the XML file is Consumer_1 and it use Logger_1 therefore we need to change LoggerType in Logger_1 section to LoggerType::Stdout.
<DefaultConsumer value="Consumer_1"/>
<ConsumerList>
<Consumer>
<Name value="Consumer_1"/>
<Channel value="Channel_1"/>
<Logger value="Logger_1"/>
<Dictionary value="Dictionary_1"/>
</Consumer>
…
<Logger>
<Name value="Logger_1"/>
<LoggerType value="LoggerType::Stdout"/>
<LoggerSeverity value="LoggerSeverity:: NoLogMsg"/>
</Logger>
