Timeout on acquiring lock when insert duplicated record in distributed mode #7391
Comments
|
In your pseudo code to reproduce the problem I don't see any edge. It's weird that you have a |
|
yes, no need edge. I just tested it on win 10, same exception. |
|
What record is #13:1 ? Can you print here the result of |
|
the result is empty, seem orientdb server catch the ORecordDuplicatedException and commit fail. |
|
additionally, there is following log in orient-server.log.0 after changing log level to FINE, seems orientdb server catch the ORecordDuplicatedException but not unlock the record.
|
|
This is interesting. So the problem is on duplicated key in the same distributed transaction? |
|
yes |
|
In branch 2.2.x we improved the management of timeouts with distributed. Could you please try with 2.2.20-SNAPSHOT if the error persists? |
|
I tried 2.2.20-SNAPSHOT from git, the code stuck at line "graph.commit()", still failed. |
|
We pushed a fix a few hours ago that seems the very same bug. Could you please build an OrientDB distribution from 2.2.x and check if this is working now? Thanks in advance. |
|
I tried to start orientdb with -Ddistributed=true in IDEA, but have following exception, did I miss something. |
|
Which release are you using? The error is |
|
I can start it using branch 2.2.18, but after checkout to 2.2.x , it fail to start . with same config. |
|
the test case passed just now, thanks. |
|
Cool, I'm closing this exception, in case the problem persists, please comment this issue, so I can reopen it. Thanks. |
|
Just to confirm. I had this same problem with Object API in distributed mode, using 2.2.20 it did not throw me this exception and halts everything, in my case it deadlocks without a timeout. After building 2.2.21 snapshot all is fine |
|
I'm using orientdb 2.22 in distribute environment, when i use concurrent thread to write the data and it occurs the exception below is my code: i will create the sql by my java code, and then using java api to execute it. java code: sql: I have some questions: Thanks |
|
I am using Orient 3.0.10 and facing same issue on concurrent write. I have created following setup -
It works well for 2 threads but as number of threads increases, it times out. Any clues? |
|
Same issue in 3.0.13. Any ideas how we can fix this? |
OrientDB Version: 2.2.18
Java Version: 1.8.0_73
OS: el6
issue:
when inserting duplicated record (with unique index) in distributed mode, have following exception:
ODistributedRecordLockedException: Timeout (3938ms) on acquiring lock on record #13:1 on server 'node2'. It is locked by request 0.14
For version 2.1.x, OrientDB will throw ORecordDuplicatedException, which I think make sense. because I can handle it immediately without delay.
why v2.2.x does not throw ORecordDuplicatedException, did I miss something?
Expected behavior
should throw ORecordDuplicatedException
Actual behavior
ODistributedRecordLockedException: Timeout (3938ms) on acquiring lock on record #13:1 on server 'node2'. It is locked by request 0.14
Steps to reproduce
create unique index:
drop class orion_v unsafe;
create class orion_v extends v;
create property orion_v.pk string;
alter property orion_v.pk mandatory true;
drop index idx_orion_v_pk;
create index idx_orion_v_pk on orion_v(pk) unique_hash_index;
run java code:
String pk1="11";
graph.addVertex("class:orion_v","pk",pk1);
graph.addVertex("class:orion_v","pk",pk1);
graph.commit();
The text was updated successfully, but these errors were encountered: