[NETCONF-256] excessive memory consumption when mounting devices Created: 11/Aug/16 Updated: 15/Mar/19 Resolved: 20/Oct/16 |
|
| Status: | Resolved |
| Project: | netconf |
| Component/s: | netconf |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Bug | ||
| Reporter: | Giles Heron | Assignee: | Unassigned |
| Resolution: | Done | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
Operating System: Linux |
||
| Attachments: |
asr9000v.txt
asr9k.txt
|
||||||||
| Issue Links: |
|
||||||||
| External issue ID: | 6391 | ||||||||
| Priority: | Highest | ||||||||
| Description |
|
Running Beryllium-SR3 on Linux I'm finding that I need to increase the max Java heap beyond 2G to mount 2 different devices running the same software version (so the models overlap - but there are some different models due to hardware specifics). Once I increased heap to 4G I was able to mount both devices, plus a third and fourth device running different software versions. At any rate this seems like excessive memory usage. Can someone please look into it. I can provide logs (too big to post here) |
| Comments |
| Comment by Jakub Morvay [ 15/Aug/16 ] |
|
How many models does these two devices provide? Can you please add them here as an attachment? Can you also provide devices schema context building parts from logs? |
| Comment by Giles Heron [ 15/Aug/16 ] |
|
The devices each provide around 400 models, but most of them overlap. |
| Comment by Giles Heron [ 15/Aug/16 ] |
|
Attachment asr9k.txt has been added with description: hello from asr9001 |
| Comment by Giles Heron [ 15/Aug/16 ] |
|
Attachment asr9000v.txt has been added with description: hello from ASR9000v |
| Comment by Giles Heron [ 15/Aug/16 ] |
|
I can send you the whole log offline? |
| Comment by Jakub Morvay [ 16/Aug/16 ] |
|
(In reply to Giles Heron from comment #5) Yeah, you can. Just write me an e-mail and we figure something out. |
| Comment by Giles Heron [ 22/Aug/16 ] |
|
Hi Jakub - any update on this? |
| Comment by Jakub Morvay [ 25/Aug/16 ] |
|
Hi Giles, I tried to look at log you sent me, but I haven't seen nothing suspicious. I will try to simulate this with test-tool with 400 models.. |
| Comment by Andrej Mak [ 20/Sep/16 ] |
|
Cause of excessive memory consumption is in yangtools. I've already created a bug. |
| Comment by A H [ 18/Oct/16 ] |
|
To better assess the impact of this bug and fix, could someone from your team please help us identify the following: |
| Comment by Andrej Mak [ 19/Oct/16 ] |
|
This issue has been resolved in master, Boron and Beryllium by fixing https://bugs.opendaylight.org/show_bug.cgi?id=6757 |
| Comment by A H [ 19/Oct/16 ] |
|
Has this bug been verified as fixed in the latest Beryllium SR4 Build 20161019? |
| Comment by Giles Heron [ 19/Oct/16 ] |
|
I can test if you point me at the build |
| Comment by Giles Heron [ 19/Oct/16 ] |
|
I can test if you point me at the build |
| Comment by A H [ 19/Oct/16 ] |
|
(In reply to Giles Heron from comment #14) |
| Comment by Giles Heron [ 20/Oct/16 ] |
|
still seeing GC overhead exceeded mounting a single XRv node with 2G of JAVA_MAX_MEM. |
| Comment by Andrej Mak [ 20/Oct/16 ] |
|
There is still high memory usage during building of schema context, so it is needed to increase heap size. When context is build, memory usage should drop to normal values. |