At what point exactly do transactions in the mempool start getting purged?

[logfiles]

I have been observing the mempool in the recent times, and it's usage has gone past 300 MBs, now getting to 500 MBs



I initially thought like 300 MBs was the limit after which the purging would start almost immediately to keep below 300 MBs but now that It's way above that, so;

1. What is the actual acceptable total size of unconfirmed transaction in the mempool before the purging begins?
2. Has the bitcoin mempool ever reached a size of 1 GB in unconfirmed transactions?

[1714210927]

1714210927

[1714210927]

1714210927

[1714210927]

1714210927

[DaveF]

1) Whatever thew node operator sets. 300 MB is the default. You can set it to whatever you want.
2) IIRC yes back in 2017 there were a few places that had no max size, or a very large one that were saying that they were above a gig.

Going to be interesting to see what happens if this keeps up. Can some of the nodes running on older more marginal hardware may not keep up.

-Dave

[pooya87]

I have been observing the mempool in the recent times, and it's usage has gone past 300 MBs, now getting to 500 MBs

Technically a mempool visualizer/analysis tool like https://mempool.space (which your screenshot seems to be from) should not purge its mempool or at the very least have a much larger cap compared to what the default is for most full nodes, so that it can give a better view of all the number of transactions and their fees.

If you want to know the real behavior of a node you should check a real node (like your own full node).

[nc50lc]

I initially thought like 300 MBs was the limit after which the purging would start almost immediately to keep below 300 MBs but now that It's way above that, so;

mempool space is probably keeping track of all unconfirmed transactions but has its own algorithm to compute the default "Purging" fee rate.
On a side note, "observing the mempool" is quite a tedious task since each node has its own unique mempool.

Additionally, for the 14day default expiration, some nodes may have set it higher so transactions may not get totally dropped after 14days.

[NotATether]

Bitcoin Core has set the 300MB mempool limit as a customizable command-line option, not as a consensus rule. That is probably why mempool.space's scale goes up to 300MB.

That means most (but not all) nodes will be evicting transactions once the total size gets above 300MB. Mempool.space is probably running a node with a very large mempool size, which is why you can still see the current size on the site.

[DaveF]

Bitcoin Core has set the 300MB mempool limit as a customizable command-line option, not as a consensus rule. That is probably why mempool.space's scale goes up to 300MB.

That means most (but not all) nodes will be evicting transactions once the total size gets above 300MB. Mempool.space is probably running a node with a very large mempool size, which is why you can still see the current size on the site.

I would like to think that most public block explorers are. We KNOW they are not, but it would be nice if they did.
Obviously if you are not showing ads or offering other services and operating out of pocket so to speak, you might not want to put $100s of dollars a month into running a public explorer. But going from 300mb to 500 or 750 should really not have a cost to you.

But, as I said before having a bloated mempool and full block after full block is going to push some marginal HW running full nodes over the edge. And IMO that is a good thing.
I have said it a few times, you CAN run a node on a RPi3 you can run a node on a 1st gen i3 with 2 GB of RAM. BUT outside of tinkering you really should not.

-Dave

[dkbit98]

I have said it a few times, you CAN run a node on a RPi3 you can run a node on a 1st gen i3 with 2 GB of RAM. BUT outside of tinkering you really should not.

Maybe that is true but it's not permanent solution for this mess that was made with ordinals and NFT spam goin on Bitcoin blockchain.
I think it would be good to somehow upgrade Taproot and introduce some limitations that could prevent things like this from happening.
Speaking from non-developer stand point, but this was obvious side effect as result of introducing Taproot fork, nobody want's rollback but we seriously need zo think about one more soft fork.

mempool space is probably keeping track of all unconfirmed transactions but has its own algorithm to compute the default "Purging" fee rate.

Timechain calendar is showing the Purge, along with all other information shownon Mempol.space website.


https://timechaincalendar.com/

[stompix]

Timechain calendar is showing the Purge, along with all other information shownon Mempol.space website.

They are using mempool.space API, so of course they have the same information as them, just shown differently.

I have said it a few times, you CAN run a node on a RPi3 you can run a node on a 1st gen i3 with 2 GB of RAM. BUT outside of tinkering you really should not.

And you could still use an old HDD while you're at it, but I'm not sure the time wasted screaming at it while doing a sync is worth the saved $. 
If building a decentralized currency aimed at toppling all world finances and revolutionizing every this and that is threatened by the price of a 100$ SSD and a bunch of monkey jpgs probably something is wrong somewhere.

[LoyceV]

If building a decentralized currency aimed at toppling all world finances and revolutionizing every this and that is threatened by the price of a 100$ SSD and a bunch of monkey jpgs probably something is wrong somewhere.

But there is no problem, right? Bitcoin works just fine, although fees went up a bit. But fees have been 100 times higher than they are now, and I don't think the meme-spammers are going to pay that much for the gigabytes they're spamming.
Bitcoin is still processing transactions. Although the full mempool and dropping price reminds me of what happened at the end of 2017.

[stompix]

But there is no problem, right? Bitcoin works just fine, although fees went up a bit. But fees have been 100 times higher than they are now, and I don't think the meme-spammers are going to pay that much for the gigabytes they're spamming.

From my point of view, no, right now we don't have a problem at all, and all those claims, including that this is "a terrorist attack" and that these fees will somehow deter, well,  transactions that were not even made first place as we had empty blocks are just exaggeration at this point. Yeah, fees have been around 400sat/b for first block inclusion before, not something I want to see again but still an order of magnitude from what we have now, so all the current panic over ordinals is, at least from my point of view, just fearmongering.

Although the full mempool and dropping price reminds me of what happened at the end of 2017.

That's the effect of Silvergate Bank closing down, not the Planet of the Apes jpegs taking over the blockchain.

[LoyceV]

That's the effect of Silvergate Bank closing down

Thanks, I didn't know that. Amazing how FUD from something I've never even heard of still works so well to make people panic sell.

[dkbit98]

But there is no problem, right? Bitcoin works just fine, although fees went up a bit. But fees have been 100 times higher than they are now, and I don't think the meme-spammers are going to pay that much for the gigabytes they're spamming.
Bitcoin is still processing transactions. Although the full mempool and dropping price reminds me of what happened at the end of 2017.

Imagine what would happen on top of bull run with much more transactions and much higher fees, it can become very hard to use Bitcoin for regular people.
Speaking about meme-spammers, they paid much higher fees on shitereum blockchain so I don't see why they wouldn't pay the same thing on Bitcoin blockchain.
Bigger problem is that someone who is malicious (including governments that have unlimited funds) can use this as a way to attack Bitcoin and blame regular people for everything.

[o_e_l_e_o]

I initially thought like 300 MBs was the limit after which the purging would start almost immediately to keep below 300 MBs but now that It's way above that

If you are using mempool.space, you can tell that they are not actually purging transactions from their mempool by the graph underneath the "Memory usage" bar which you posted. At the moment it states it is purging anything below ~2 sats/vbyte, and yet the graph still shows almost 40 MvB of transactions paying between 1 and 2 sats/vbyte.

Compare this to Johoe's mempool here: https://jochen-hoenicke.de/queue/#BTC%20(default%20mempool),8h,weight.
He is showing only around 3 MvB of transactions paying between 1 and 2 sats/vbyte. It seems he is indeed purging at around 300 MB of memory use.

[DaveF]

That's the effect of Silvergate Bank closing down, not the Planet of the Apes jpegs taking over the blockchain.

And now with Silicon Valley Bank getting shut down it is now causing pressure on USDC so now more people will be going back to BTC which will fill up the mempool even more as people try to move their USDC out. Personally I am wondering if I should gamble a bit and buy some in case it goes back up on Monday.

Wonder how many people are panic selling / moving coins and if it will recover on Monday.

-Dave

[stompix]

That's the effect of Silvergate Bank closing down, not the Planet of the Apes jpegs taking over the blockchain.

And now with Silicon Valley Bank getting shut down it is now causing pressure on USDC so now more people will be going back to BTC which will fill up the mempool even more as people try to move their USDC out.

I don't think so, people that held USDC were doing so for trading and to move funds easily around exchanges, Bitcoin and self-custody aren't what they were looking for, and even if they do buy coins on the exchanges I doubt they will take them out. The last block had 2sat/b transactions and I don't see anything unusual in terms of incoming transactions volume that is not ordinals, exchanges wouldn't go and batch transactions with 1sat/b  fees either.
Of course, we're also 20 blocks over the normal volume for 24h but I still doubt there will be another spam wave.