< bitcoin-git>
bitcoin/0.14 569596c Suhas Daftuar: Don't require segwit in getblocktemplate for segwit signalling or mining...
< bitcoin-git>
bitcoin/0.14 2cd2cd5 Suhas Daftuar: Test transaction selection when gbt called without segwit support...
< bitcoin-git>
bitcoin/0.14 ccb47bf Wladimir J. van der Laan: Merge #10006: [0.14 backport] Don't require segwit in getblocktemplate for segwit signalling or mining...
< bitcoin-git>
[bitcoin] practicalswift opened pull request #10027: Set to nullptr after delete (master...set-to-nullptr-after-delete) https://github.com/bitcoin/bitcoin/pull/10027
< bitcoin-git>
[bitcoin] practicalswift opened pull request #10029: Fix parameter naming inconsistencies between .h and .cpp files (master...fix-header-parameter-naming-inconsistencies) https://github.com/bitcoin/bitcoin/pull/10029
< adiabat>
OK I don't feel like as much of an idiot; the 0.14.0 bitcoind crashed again
< sipa>
OOM>
< sipa>
?
< adiabat>
this is with dbcache=64 and mempool=32
< sipa>
ouch
< sipa>
how much ram?
< adiabat>
I don't think it's OOM; I don't get any St9bad_alloc or anything like that
< adiabat>
this is 1GB, a raspberry pi 2
< sipa>
0.14 will just exit in the case of an OOM
< adiabat>
ok, could still be OOM
< adiabat>
what debug logging should I enable to see this better?
< sipa>
look at dmesg?
< sipa>
ah, no
< sipa>
nvm
< adiabat>
ah
< adiabat>
actually, yes!
< sipa>
0.14 will try to write to debug.log that it ran out of memory
< adiabat>
Out of memory: Kill process 13720 (bitcoind) score 333 or sacrifice child
< sipa>
ah
< sipa>
1 GB is very little
< adiabat>
whole bunch of stuff in dmesg
< sipa>
plus RPi loses some memory for the GPU
< adiabat>
yeah, could set up swap I guess? This is testnet though... although I guess there's no difference :)
< sipa>
swap would move it from crashing to unusably slow :)
< adiabat>
well, perhaps, but it seems to last a day or two without swap
< adiabat>
so maybe it would only use swap occasionally
< adiabat>
I haven't actually used swap in linux for a long time though
< adiabat>
super hack-y but I might just make a watchdog script that restarts it if it's dead
< btcdrak>
the 2GB pine64 needs swap to run bitcoind remotely well.
< sipa>
i'm trying to fix the memory peak during flush, but it seems nontrivial
< sipa>
it's been worsened in 0.14 due to the borrowing of mempool
< gmaxwell>
adiabat: that may just be another issue.
< gmaxwell>
adiabat: can you run bitcoind inside gdb to see if you can learn more of the crash? gdb ./bitcoind <enter> then run -daemon=0 <enter>
< adiabat>
gmaxwell: will run that and leave it; might take a while to crash
< gmaxwell>
if it crashes, type "thread apply all bt" and save the couple screenfulls of output.
< adiabat>
will do
< adiabat>
also wondering if it may be arm-specific; I think there are a decent number of testnet nodes on little VMs like digital ocean w/ 1GB