Depends on D3582
We should see a proper status in the terminateCallback and the error log(if an error occurs).
Details
Details
cd services yarn run-backup-service
Diff Detail
Diff Detail
- Repository
- rCOMM Comm
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
Comment Actions
How do we plan to verify the data?
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | Why do we need to change the order here? It looks like the only difference is that we will write the error every time this method is called instead of once |
Comment Actions
I was planning to check if the hash is valid, we can also send a bool flag manually at the end as suggested in D3562. Anyway, relying on !ok here is a no-go.
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | We need that because terminate is called from several places, so we should assume there may be a situation when it's called twice and an error occurs between these two calls. So we have something like this: terminate(OK); ... terminate(Err); Now, we want to have that error printed out, right? Of course, there may be a situation when it gets called twice and there's an error for both calls: terminate(Err); ... terminate(Err); In such a case, we will have this error printed twice. If that bothers you, I can add a bool flag preventing that. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | This is confusing. What is the scenario in which terminate is called twice? In this function, we set done to true, so after it is set, there shouldn't be any operation that can fail. Or maybe, the meaning of done is different and we should find a better name. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) |
It should never happen, but we can assume it will somehow. There's no hard check for this, so maybe there's some edge case for this.
The point is to not invoke StartWritesDone multiple times. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | So when this->done we should add a warning Terminate function was called multiple times, which should never happen. Please investigate! |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | Yes, as I suggested in D3582, we should probably throw in such a case. That would terminate the connection with an error and we could just set it to an internal server error. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | What would be the consequence of calling terminate twice? If we can continue execution I'd prefer to just log an error. But if it is not possible or it is likely to cause some serious issues, it's safer to throw. Overall, it is better to recover from such issues instead of always throwing. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 56–61 ↗ | (On Diff #10879) | Ok so let's just log. |
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 55 ↗ | (On Diff #11331) | yes |
| 56–61 ↗ | (On Diff #10879) | I spent some time on this and I think that terminate will be called twice when for example there are no more reads:
And this is just fine. I think we should allow it. What we should prevent is overwriting the error status. So if there's an error and then we somehow call terminate with the status OK, we should not suppress the error. I'm going to push changes for this to reduce cycles, I think this is a valid change. If you feel otherwise, I can always revert. |
Comment Actions
Do we need to introduce validate just like in D3582?
| services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h | ||
|---|---|---|
| 57–59 ↗ | (On Diff #11383) | Shouldn't we validate before telling that there was an error? Or is this case different than D3582? |
| 56–61 ↗ | (On Diff #10879) | Agree, it's a good practice to make such changes to reduce cycles. Regarding terminate - sounds reasonable. |