change checkpoint_ns to 255 (temp fix) #40
Conversation
| @@ -119,6 +119,7 @@ def setup(self) -> None: | |||
| ): | |||
| cur.execute(migration) | |||
| cur.execute(f"INSERT INTO checkpoint_migrations (v) VALUES ({v})") | |||
| cur.execute("COMMIT") | |||
There was a problem hiding this comment.
Do you know why the upstream PostgreSQL checkpointer doesn't need this? (I assumed it was since it expects connections to be constructed using autocommit=True, but I'm open to making this support other cases too if I understand the use case.)
There was a problem hiding this comment.
In my testing, the version value is not up to date. So when I run setup next time. It will try to excute previouse migration. Previous migration will cause error.
I didn't dive much too. But I found execute 'commit' can make version up to date. So I added.
And now setup() can been call multiple times.
There was a problem hiding this comment.
Maybe it's caused by autocommit . When I use this repo. I read the readme. Noticed that I need autocommit set to True.
But when I use sqlalchemy import create_engine. I didn't find autocommit setting. After searching, I think the autocommit is set to True by default. So I didn't set autocommit in my code. Just use default value.
|
@tjni en. 255 is not enough. As I use nested subgraph. And I found that one nested subgraph call will use 40-60 chars for checkpoint-ns. So after about 5 nested subgraph call will cause this issue again. So temprarly I just avoid too much nested subgraph. But I think about 5 nested subgraph call is not enough for large project But I'm not testing postgresql, if it also have this limitation or not? As I'm using MySQL in my project now. |
|
Let's update the primary key then and see if that fixes things! |
|
I pushed migrations to remove |
|
I released version 2.0.9 with the removal of checkpoint_ns from the primary keys. Please let me know if there are any issues. |
|
Thank you for the quick reply and fix. It looks work now. Thanks very much. |
|
FYI: I released version 2.0.10 which adds back a hash of checkpoint_ns to the primary key. It turns out that this is used in the upsert statements. |
try fix #39
but mysql don't support TEXT as a key. this MR is just a temp fix.
for long term fix. I think we should remove checkpoint_ns from primary key