In the past several hours there have been at least two blocks with a sufficient number of transactions per block to leave bitcoin nodes relying on Berkeley Database for block handling to wedge when set to the post March 2013 limit of 40,000 database locks and objects. For a few hours doubling that amount to 80,000 sufficed until a still more complex block arrived. A patch has recently been published which should remedy this issue until such a time the universe undergoes heat death. The patch works by raising the maximums Berkeley Database is configured with in order to handle any number of transactions that can fit into a Bitcoin block. On some platforms like OpenBSD which aggressively allocate memory in advance for safety reasons Bitcoin's RAM usage is increased noticeably with this patch. If your system enforces low per-process memory limits you may have to edit you system's settings.
 |
Qntra - Herald Of The Most Serene Republic |
Does this impact the Bitcoin Core releases that use BSD as well?
It affects all pre-0.8 versions of Bitcoin and their forks which still use BDB to handle block processing. The BSD note was that on OpenBSD the patch increases RAM usage. Linux users aren't reporting substantially increased RAM usage. The difference is likely in the ways they operate memory.
This only affects ancient, unsupported versions of Bitcoin Core that are not safe to use even unrelated to this.
Also, the "post March 2013 limit" (included in 0.4.9rc2 and 0.5.8rc2) was 537k, not 40k – so this problem never existed in the first place.
The product in question, TheRealBitcoin, was patched – and the patch applied to two infrastructural nodes – within two hours of the wedge having been discovered.
If this is "unsupported", I should like to learn what "supported" looks like.
"Bitcoin Core" is the name a collection of USG moles and assorted scammers are giving to a pile of spaghetti that's unfit to run on Windows even.
It did enjoy some temporary attention in the early days of Bitcoin, chiefly through the happenstance that there not being any actual brains involved with the project.
Meanwhile it's lost any importance, either intellectual or practical. Miners run homebrew, thinking people run the Bitcoin Foundation client. That's all, really.