[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17509834#comment-17509834 ] Josef Zahner commented on NIFI-8435: Just for reference. We have upgraded from NiFi 1.13.2 to NiFi 1.15.3 and adapted the property "Kudu Client Worker Count" to the number of "Concurrent Tasks" accordingly. So far no leaks or high memory situations anymore. However it's not ideal to have the property "Kudu Client Worker Count" as high as the number of CPUs as we have bladeservers with 64 CPUs and about 30 PutKudus in place -> leads to a lot of kudu client workers. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Fix For: 1.14.0 > > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1.5h > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.20.1#820001)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17348169#comment-17348169 ] Josef Zahner commented on NIFI-8435: All right, we try to test the 1.14.0 snapshot build, but it will take a few days as my colleague is in vacation. I'll post the result when we have finished the test. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Fix For: 1.14.0 > > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1.5h > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17348128#comment-17348128 ] Pierre Villard commented on NIFI-8435: -- In the configurations [~pgyori], [~exceptionfactory] and I tested, we think we didn't have a memory leak leading to OOME. However we did observe a large number of threads being instantiated by Netty and we noticed that the default number would be a factor of the number of cores on the host. That's why we think that if having many instances of the processor, with many concurrent tasks and when running on nodes with a lot of cores, this can cause issues. We think that this change is enough because it gives the option to override this default value. There is still the open Netty issue linked in this JIRA that we are going to monitor closely. To be honest, if you have the chance to test this change in one of your environments, that would definitely be a valuable feedback and we can re-open the JIRA if the situation didn't improve. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Fix For: 1.14.0 > > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1.5h > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17348073#comment-17348073 ] Josef Zahner commented on NIFI-8435: Thanks [~pvillard], did you do some tests to verify your change? How big is the chance that the issue has been fixed by your change? > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Fix For: 1.14.0 > > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1.5h > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17347839#comment-17347839 ] ASF subversion and git services commented on NIFI-8435: --- Commit 13e5dc4691c8d6f7fb8fafe6ed16caeb603cd315 in nifi's branch refs/heads/main from David Handermann [ https://gitbox.apache.org/repos/asf?p=nifi.git;h=13e5dc4 ] NIFI-8435 Added Kudu Client Worker Count property - Implemented custom ThreadPoolExecutor with maximum pool size based on Worker Count property - Refactored processing methods to ensure KuduSession is always closed - Added SystemResourceConsideration for Memory - Removed duplicate dependency on nifi-security-kerberos - Adjusted method naming to clarify functionality - Reverted addition of defaultAdminOperationTimeoutMs() Signed-off-by: Pierre Villard This closes #5020. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1h 20m > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17335560#comment-17335560 ] David Handermann commented on NIFI-8435: [~granthenke] might have a better handle on whether there are any particular features or fixes necessary in the newer version of Netty. However, 4.1.43.Final is over a year old and recent versions have included a large number of bug fixes, so it may not be a simple solution. Another potential workaround is for the Kudu client to expose the allocator as a configurable option, which would NiFi PutKudu to set a different option and avoid the issue with the pooled allocator. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1h 20m > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17335553#comment-17335553 ] Josef Zahner commented on NIFI-8435: I don't understand the details, but wouldn't it be a quick fix to force PutKudu to use the netty version from NiFi v.1.11.4 in the main pom? Or are there any features required from the new netty version? > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1h 20m > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17335496#comment-17335496 ] David Handermann commented on NIFI-8435: For reference, there is an open issue with Netty, starting in version 4.1.43.Final that appears to have very similar behavior as the issue observed with PutKudu 1.13.2. The comments on the issue mention several potential workarounds, but no great options for the purposes of the PutKudu processor. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > Time Spent: 1h 20m > Remaining Estimate: 0h > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[
https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17326985#comment-17326985
]
David Handermann commented on NIFI-8435:
[~jzahner] I put together PR 5020 to add a {{Kudu Client Worker Count}}
property for {{PutKudu}}. The default value uses the same approach as
{{KuduClient}}, but for your use case, it seems like a smaller number, matching
the number of Concurrent Tasks configured for PutKudu, would be better. The PR
also ensures that {{KuduSession}} is always closed. The custom worker count
number also provided the opportunity to set a custom ThreadFactory so that
thread names now include the Processor Identifier, which should be helpful for
troubleshooting. Kudu changes from Netty 3 to Netty 4 have definitely impacted
the memory footprint, so I do not expect these changes to be a complete
resolution. However, if you have any feedback on the PR, feel free to add
comments.
> PutKudu 1.13.2 Memory Leak
> --
>
> Key: NIFI-8435
> URL: https://issues.apache.org/jira/browse/NIFI-8435
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
>Affects Versions: 1.13.2
> Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu
> 1.10.0
>Reporter: Josef Zahner
>Assignee: Peter Gyori
>Priority: Critical
> Labels: kudu, nifi, oom
> Attachments: Screenshot 2021-04-20 at 14.27.11.png,
> grafana_heap_overview.png, kudu_inserts_per_sec.png,
> putkudu_processor_config.png, visualvm_bytes_detail_view.png,
> visualvm_total_bytes_used.png
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with
> PutKudu.
> PutKudu on the 1.13.2 eats up all the heap memory and garbage collection
> can't anymore free up the memory. We allow Java to use 31GB memory and as you
> can see with NiFi 1.11.4 it will be used like it should with GC. However with
> NiFi 1.13.2 with our actual load it fills up the memory relatively fast.
> Manual GC via visualvm tool didn't help at all to free up memory.
> !grafana_heap_overview.png!
>
> Visual VM shows the following culprit: !visualvm_total_bytes_used.png!
> !visualvm_bytes_detail_view.png!
> The bytes array shows millions of char data which isn't cleaned up. In fact
> here 14,9GB memory (heapdump has been taken after a while of full load). If
> we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a
> few hundred MBs.
> As you could imagine we can't upload the heap dump as currently we have only
> productive data on the system. But don't hesitate to ask questions about the
> heapdump if you need more information.
> I haven't done any screenshot of the processor config, but I can do that if
> you wish (we are back to NiFi 1.11.4 at the moment).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17326617#comment-17326617 ] David Handermann commented on NIFI-8435: Thanks for the confirmation [~jzahner]. In attempting to reproduce the issue, I observed that the each KuduClient instance can spawn a number of threads in some circumstances. On a small NiFi instance, stopping the PutKudu processor temporarily increases the number of threads. More recent versions of the Kudu client library incorporated an upgrade from Netty 3 to Netty 4, which included changes to how Netty handles pooled byte buffering. Initial testing with Netty leak detection enabled did not yield any definite findings, but memory usage is clearly connected directly to the number of PutKudu Processor instances, and the associated KuduClient instances. The KuduClient supports configuring the number of worker threads, so this may be an area of improvement, although it does not appear to be a complete answer to the problem. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17326351#comment-17326351 ] Josef Zahner commented on NIFI-8435: [~exceptionfactory] yes I have a list. My heapdump which I'm referencing to has about *1000* "kudu-nio-xxx" threads running. The whole heap dump has a size of about 15 GB. I've also a heap dump which I've created shortly after a fresh start of NiFi and it shows only about 120 "kudu-nio-xxx" threads, size is only 2.5 GB. So you see the number of threads is growing over time. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17326053#comment-17326053 ] David Handermann commented on NIFI-8435: [~jzahner] Do you have the list of running threads available from the heapdump you referenced? The Kudu client spawns a number of threads prefixed with "kudu-nio", so it would be helpful to know how many were running when you hit the memory issue. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325761#comment-17325761 ] Josef Zahner commented on NIFI-8435: voilà, you got it ;). that's exactly what we see. You can press "GC" in visualvm to force garbage collection, if it doesn't go back to the lowest value on the left hand side you have for sure a leak. If you create a heap dump you will probably see that PutKudu leaks. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325751#comment-17325751 ] Peter Gyori commented on NIFI-8435: --- Ok, with greater load I see: !Screenshot 2021-04-20 at 14.27.11.png! at the bottom of the spikes, the memory usage does not go back to the initial ~200MB in this case, there is a slight increase each time. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: Screenshot 2021-04-20 at 14.27.11.png, > grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325717#comment-17325717 ] Josef Zahner commented on NIFI-8435: Ok perfect, I can see the sawtooth as well until the heap is exhausted (and the sawtooth gets smaller as more the memory has been filled up). Please have a look at the heap "byte" objects, at least if you generate enough flows/records. I don't think that 3 small records every 0.1 seconds are enough. In my initial grafana screenshot, where you see the heap size, we were inserting about 600K records per second, just that you get a rough estimation of our workload and how long it took to see the issue. !kudu_inserts_per_sec.png! > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, kudu_inserts_per_sec.png, > putkudu_processor_config.png, visualvm_bytes_detail_view.png, > visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325691#comment-17325691 ] Peter Gyori commented on NIFI-8435: --- Hi [~jzahner] , I was (and still am) trying to reproduce the issue. Yes, I'm using VisualVM and checked the heap dump. I have a 3-node cluster with Kudu, and NiFi is running on my computer with 1 PutKudu processor, 2 threads and inserting 3 small records every 0.1 seconds into Kudu. It is not as heavy load as yours but in VisualVM I can still see the sawtooth diagram with GC freeing the memory every ~1 minute or so. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, putkudu_processor_config.png, > visualvm_bytes_detail_view.png, visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325679#comment-17325679 ] Josef Zahner commented on NIFI-8435: Sorry, I missed to ask. But how did you checked the memory leak? Did you analysed the HEAP dump as I've done it with visualvm? Of course it happens only if you produce data/flows and if you have just a little bit of data you won't see the issue that fast without a tool to analyse the heap dump. We are inserting overall a few GB per minutes to kudu in production. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, putkudu_processor_config.png, > visualvm_bytes_detail_view.png, visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325677#comment-17325677 ] Josef Zahner commented on NIFI-8435: [~pvillard] I've added the processor config below. We can clearly reproduce the memory leak issue within our lab with the same config on a 4-node NiFi cluster. Yes we are using multiple concurrent tasks and on our actual config we have about 34 PutKudu processors running in parallel (with multiple threads). !putkudu_processor_config.png|width=1042,height=729! As emergency fix we have built the NAR from 1.11.4 for NiFi 1.13.2. It runs since a few hours and no leak so far. But of course we would like to use the official PutKudu and not an old version of it. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, putkudu_processor_config.png, > visualvm_bytes_detail_view.png, visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17325241#comment-17325241 ] Pierre Villard commented on NIFI-8435: -- [~jzahner] - can you share the configuration of your processor? Are you using multiple concurrent tasks? what's your configuration in terms of properties? Anything you can share would be useful as I've not been able to reproduce the issue after some attempts. > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Assignee: Peter Gyori >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, > visualvm_bytes_detail_view.png, visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and faced a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (NIFI-8435) PutKudu 1.13.2 Memory Leak
[ https://issues.apache.org/jira/browse/NIFI-8435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17322186#comment-17322186 ] Josef Zahner commented on NIFI-8435: [~granthenke] hope you can help here. The processor has the same config as shown in my old bug report here (of course the new properties are missing): https://issues.apache.org/jira/browse/NIFI-6908 > PutKudu 1.13.2 Memory Leak > -- > > Key: NIFI-8435 > URL: https://issues.apache.org/jira/browse/NIFI-8435 > Project: Apache NiFi > Issue Type: Bug > Components: Extensions >Affects Versions: 1.13.2 > Environment: NiFi 1.13.2, 8-Node Cluster running on CentOS 7, Kudu > 1.10.0 >Reporter: Josef Zahner >Priority: Critical > Labels: kudu, nifi, oom > Attachments: grafana_heap_overview.png, > visualvm_bytes_detail_view.png, visualvm_total_bytes_used.png > > > We just upgraded from NiFi 1.11.4 to 1.13.2 and face a huge issue with > PutKudu. > PutKudu on the 1.13.2 eats up all the heap memory and garbage collection > can't anymore free up the memory. We allow Java to use 31GB memory and as you > can see with NiFi 1.11.4 it will be used like it should with GC. However with > NiFi 1.13.2 with our actual load it fills up the memory relatively fast. > Manual GC via visualvm tool didn't help at all to free up memory. > !grafana_heap_overview.png! > > Visual VM shows the following culprit: !visualvm_total_bytes_used.png! > !visualvm_bytes_detail_view.png! > The bytes array shows millions of char data which isn't cleaned up. In fact > here 14,9GB memory (heapdump has been taken after a while of full load). If > we check the same on NiFi 1.11.4, the bytes array is nearly empty, around a > few hundred MBs. > As you could imagine we can't upload the heap dump as currently we have only > productive data on the system. But don't hesitate to ask questions about the > heapdump if you need more information. > I haven't done any screenshot of the processor config, but I can do that if > you wish (we are back to NiFi 1.11.4 at the moment). -- This message was sent by Atlassian Jira (v8.3.4#803005)
