with mysql, duplicate entries cause failures unlik...
# spicedbu
_definitely_not_sam_
06/21/2022, 2:06 PMwith mysql, duplicate entries cause failures unlike with postgres where they are ignored. is this intentional?
v
vroldanbet
06/21/2022, 2:07 PM👋🏻 Hi! Do you mind elaborating? What operation are you performing, and what is your expectation?
u
_definitely_not_sam_
06/21/2022, 2:09 PMin the postgres version if i wrote a duplicate relation it would perform the sql equivalent of an "insert ignore". so if it was a duplicate it silently continued whereas now it fails the write operation with:
Error 1062: Duplicate entry 'client-CLIENT_ID-admin-user-USER_ID' for key 'uq_relation_tuple_living'_definitely_not_sam_
06/21/2022, 2:10 PMthis occurs on
WriteRelationships_definitely_not_sam_
06/21/2022, 2:12 PMfull disclosure, I am using TiDB not vanilla MySQL
v
vroldanbet
06/21/2022, 2:14 PMare you doing
CREATETOUCHvroldanbet
06/21/2022, 2:17 PMand further question: are you attempting to write the duplicate:
- in 2 write RPCs
- as part of one write RPC with multiple tuples, where some are duplicated
u
_definitely_not_sam_
06/21/2022, 2:18 PM1. Create, I wan't aware of touch until just now
2. The duplicate exists already. This is a bootstrap script that died half way through last time
v
vroldanbet
06/21/2022, 2:20 PMso esentially you are doing:
- write(tupleA)
- write(tupleA)
instead of:
- write(tupleA, tupleA,...)
correct?
u
_definitely_not_sam_
06/21/2022, 2:21 PMyes. let me try with the touch. I am updating spice for the first time in ~6 months so i don't think touch was a thing last time i worked on it.
v
vroldanbet
06/21/2022, 2:24 PMwell, based on your comments, it would seem like there is misalignment between MySQL and PostgreSQL. You are supposed to get an error.
u
_definitely_not_sam_
06/21/2022, 2:24 PMtouch allows upsert correctly
v
vroldanbet
06/21/2022, 2:24 PMright
u
_definitely_not_sam_
06/21/2022, 2:24 PMdid touch exist 6mo ago?
v
vroldanbet
06/21/2022, 2:25 PMcan you confirm PSQL does not fail with duplicate tuples with`CREATE` ?
u
_definitely_not_sam_
06/21/2022, 2:25 PMOne sec
_definitely_not_sam_
06/21/2022, 2:26 PMCREATE performs an upsert as of 7mo ago, using docker image
730ff726f044_definitely_not_sam_
06/21/2022, 2:27 PMThe spice system hasn't been updated in awhile so atm i cannot confirm on the latest release
v
vroldanbet
06/21/2022, 2:28 PMgot it! I appreciate the info 🙇🏻♂️ There should be in theory tests to demonstrate
CREATEu
_definitely_not_sam_
06/21/2022, 2:28 PMOnce I complete the update to v1.8.0 and move to TiDB I will be able to confirm on postgres
v
vroldanbet
06/21/2022, 2:28 PMthat'd be awesome, thank you!
vroldanbet
06/21/2022, 2:28 PMwe would love to hear your experience with TiDB
u
_definitely_not_sam_
06/21/2022, 2:28 PMthat could be as early as later today of closer to the end of the week
v
vroldanbet
06/21/2022, 2:28 PMso please let us know!
u
_definitely_not_sam_
06/21/2022, 2:29 PMI'm a TiDB fanboy so I'm sure it will be nothing but rainbows and butterflies
v
vroldanbet
06/21/2022, 2:30 PMlol I really hope that's the case 😛 I worked with my team back at GitHub on the MySQL implementation, and I'm aware there is always something slightly different across implementations of the MySQL wire protocol
u
_definitely_not_sam_
06/21/2022, 3:38 PMOn TiDB or Spice?
v
vroldanbet
06/21/2022, 3:39 PMActually referring to Vitess
u
_definitely_not_sam_
06/21/2022, 3:59 PMVery cool! Idk if you've played with TiDB much but if you get a chance, give it a whirl. It's the most stable system I've ever used. It's self healing capabilities outside of kube are unlike any system I've ever used
v
vroldanbet
06/21/2022, 4:07 PMyeah I had a look at the architecture a while back and it's awesome. Love how's architected to support OLTP and OLAP workloads 😍
3 Views