Github user yuanchenl commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
Right...
Though it takes awhile to copy the data between nodes, it's probably still
faster for the low overhead.
I'll test #337 ASAP.
---
Github user zuyu commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
Distributing the tbl files to every data node is a better approach for
large data files. On the other hand, `InsertTuples` has a lot overhead by
attribute type checks and tuple copy costs.
Github user yuanchenl commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
Yep. It was built again in docket container.
About the testing, I wrote a script that parses the tbl into insertion
command. And currently Iâm testing using this workaround.
Github user zuyu commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
@yuanchenl #337 should avoid the segfault. But the root cause of the
segfault is that we do not have a distributed FS for the tbl files. A
workaround is that copying all tbl files to every
Github user yuanchenl commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
The seg fault of the executor happens when copy command is used.
But I didn't encounter the seg fault when I am using regular insertion
---
Github user yuanchenl commented on the issue:
https://github.com/apache/incubator-quickstep/pull/336
Sorry for the late reply.
Seg fault encountered when I am trying to copy content from tbl to an
existing table.
Here's the output of the executor: