On 12/15/2015 10:16 AM, mlmikael wrote:
> Ah - so setting a node's mempool to zero [size], will make the node
> receive and process blocks, but all transactions that are not in a block
> will be ignored?

This is the size of the circular mempool buffer, which if zero will
prevent retention of txs (not block-associated):

https://github.com/libbitcoin/libbitcoin-server/blob/master/data/bs.cfg#L102

This is the transaction relay flag that gets set in the P2P version
message, indicating whether the peer should relay transactions:

https://github.com/libbitcoin/libbitcoin-server/blob/master/data/bs.cfg#L33

e


> (I.e. exactly what I asked for - if so great!)
>
> On 2015-12-15 14:33, Eric Voskuil wrote:
>> Message backlogging during sync is the cause of excessive (non-linear)
>> memory consumption (and eventual process termination). The mempool is
>> not an issue.
>>
>> But you can control both the sending of txs by (cooperative) peers and
>> the size of the mempool. Setting a zero-sized mempool in configuration
>> also instructs peers to not relay transactions.
>>
>> e
>>
>> On 12/14/2015 06:28 AM, mlmikael wrote:
>>> (After January when the sync bug will be resolved,)
>>>
>>> The only thing that can make a node's RAM consumption skyrocket (as in
>>> increase 5x above average or beyond) would be transactions, right?
>>>
>>> To that end, if I wanted a node with stability guarantees, could I just
>>> make it not listen to transactions (not relaying them is fine!) but
>>> instead just record blocks?
>>>
>>> (Or, store transactions to persistence, intentionally knowing that that
>>> will sink performance but guarantee stability?)
>>>
>>> That way, you get like a "blockchain database server".
>>>
>>> Thanks
>>>
>>> _______________________________________________
>>> Libbitcoin mailing list
>>> Libbitcoin@???
>>> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/libbitcoin
>