Never follow bitcoind backwards #3274
Conversation
tests/test_misc.py
Outdated
| sync_blockheight(bitcoind, [l1]) | ||
| l1.stop() | ||
|
|
||
| # Now shrink chain (invalidateblock leaves 'headers' field) |
There was a problem hiding this comment.
Just tried it myself, and indeed the headers remains unchanged after invalidateblock, however if invalidateblock is followed by a restart headers will reset to the expected value. Maybe that's a more stable way of going backwards than blowing away the underlying directory?
There was a problem hiding this comment.
Ah, that would have been easier!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
rustyrussell
force-pushed
the
guilt/bitcoind-never-backwards
branch
from
e09d68d
to
528b975
Compare
|
Rebased to use the Decker technique for creating reorgs... |
| time.sleep(5) | ||
| assert l1.rpc.getinfo()['blockheight'] == 110 |
There was a problem hiding this comment.
This could be made into a wait_for. Happy to fix it in a cleanup though.
There was a problem hiding this comment.
No, it can't: we want to make sure it's still at 110.
|
Seems we have a bit of collateral damage:
Other than that LGTM 👍 |
rustyrussell
force-pushed
the
guilt/bitcoind-never-backwards
branch
from
528b975
to
7cbfbc4
Compare
Thanks, fixed. I actually like the change in behavior (refusing to start) if you ask it to rescan from a not-yet-present block, so I've kept and tested that. |
rustyrussell
force-pushed
the
guilt/bitcoind-never-backwards
branch
from
7cbfbc4
to
10a74c7
Compare
This leads to all sorts of problems; in particular it's incredibly slow (days, weeks!) if bitcoind is a long way back. This also changes the behaviour of a rescan argument referring to a future block: we will also refuse to start in that case, which I think is the correct behavior. We already ignore bitcoind if it goes backwards while we're running. Also cover a false positive memleak. Changelog-Fixed: If bitcoind goes backwards (e.g. reindex) refuse to start (unless forced with --rescan). Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
rustyrussell
force-pushed
the
guilt/bitcoind-never-backwards
branch
from
10a74c7
to
cdafcf3
Compare
|
Ack cdafcf3 Fixed all the test collateral damage. |
This solves the underlying reason why #3243 was such a problem!