Domain renewed twice in one request

[vohmar]

registrar run the renew request once, but it looks like registry run the update twice. As a result domain expire date was extended double the period, registrar credit was reduced only for one.

few issues that pop out. renew request requires specifying current exp date for sanity check and to avoid such unwanted double operations so the update request no2 should have failed, but did not (race condition?). Credit management - two successful renews should have resulted in double the cost.

anonymized request:

2brenewed.ee 2021-08-21 1 1234

response:

Command completed successfully 2brenewed.ee 2023-08-21T00:00:00+03:00 1234 ccReg-6543

from the system log we see that system run the domain update twice:

Jul 21 16:06:03 reg-l-epp2 reg.internet.ee[26534]: [5c88239f-c32e-4e03-b1d4-588809c91de8] [1.2.3.4] Epp::Domain Update (0.9ms) UPDATE "domains" SET "valid_to" = $1, "updated_at" = $2 WHERE "domains"."id" = $3 [["valid_to", "2022-08-20 21:00:00"], ["updated_at", "2021-07-21 13:06:03.280056"], ["id", 82542]]

Jul 21 16:06:03 reg-l-epp2 reg.internet.ee[26534]: [5c88239f-c32e-4e03-b1d4-588809c91de8] [1.2.3.4] Epp::Domain Update (1.5ms) UPDATE "domains" SET "valid_to" = $1, "updated_at" = $2 WHERE "domains"."id" = $3 [["valid_to", "2023-08-20 21:00:00"], ["updated_at", "2021-07-21 13:06:03.494434"], ["id", 82542]]

[virtualfeline]

Considering that this resulted from a renewal request that was submitted only once I posit that credit management was at this point, actually, working correctly. I'm not sure we would want to add insult to injury and have system start charging double credits for single requests whenever something goes wrong...

[vohmar]

just purely from technical perspective it is also a problem if renew was run twice or domain was renewed for 2y but charged only for one. We fix bugs here , nothing todo with insulting or injuring anyone. Lets stay on the point, try to figure it out, but most importantly make sure it does not happen again.

[yulgolem]

Well, actually that might me an issue from my side (backend), related to some twisted logic - looks like at least background jobs are out of the scope atm.
I'll take a look on it on Monday, since I got some logs.