Couple of different things.
In the Oval Room meeting, the man in the crowd who criticised Z for not wearing suit was none other than BF of Marjorie Green Taylor - she’s a full-on MAGA fruitcake, Donald to the core. I suspect he was a plant.
My theory before all of this was D would arrange an impossible “peace treaty”, then blame Z for not signing, then withdraw all US aid.
News now coming out aid is now being “considered” and also sanctions against RU being “examined to see which could be removed”. Donald is now calling for, and having all his fully controlled appointed men ask for, Z to resign. He is voice in wilderness with this. Reds of world horrified and incredulous.
You’ll note in fact the absence of any kind of treaty at all, flat out - there isn’t even a pretence of one. I was wrong about that. What’s happening is D has completely surrendered, from a position of absolute strength, to Vladimir, and is directing US policy to the benefit of RU, in exchange for zero, nothing.
I think D is in some way beheld to V.
So we have nothing for a treaty - not even a shame cease-fire treaty - and we will now see D cut of all aid, and remove sanctions, and push for Z to resign. Also of course with this already we cannot in any way think to rely on US defence of Europe itself, so that’s already happened, whether or not the troops have left yet.
Now here’s something new and interesting. Talking to business acquaintances today, West Coast USA, they are hearing allegations that Biden had a minerals already with Z, and Z said he would sign and then did not. I strongly think this to be complete fabrication.
So where does this leave us?
In more or less exactly situation expect. USA is out of the game. I didn’t think about sanctions, but obviously good for V to have them removed. Only wildcard now is if USA actively, not just passively, works with RU - can EU continue to buy munitions from USA?
EU is now stepping up. Seizure of RU asserts on the cards, large scale defence increases, even boots on ground and planes in air, in UA, being talked about. EU finally with the picture. EU forming new group, without USA, Hungary, and Slovakia, so it will be functional.
Donald I would say has overplayed his hand. He’s not removed Z, but D has revealed himself. EU now under no illusions and is now getting its act together.
EU economy basically equal to US economy in size. EU is strong, if it goes to 4% defence spending, and it can.
Need time to re-arm though - first two years are slow. Boots on ground in UA needed during that time to keep UA up.
RU economy looks now to have peaked - just starting to begin to weaken. Given the massive costs of war, I think they have a long way to drop - they’ve been like a man throwing back Red Bull to keep going.
US release of sanctions will help that, but won’t stop it.
The coup in USA proceeds - judges and courts on the way - and corruption is kicking in. The bitcoin reserve is a way to channel vast amounts of money to cronies. Tariffs are going to totally fuck the economy. Venezuela in the USA now.
USA needs a revolution, but I think the problem is they, like everyone to whom this happens, can’t believe it’s actually happening.
By the time they really understand, it will need mass violence on the streets of WDC to eject Donald.
My American friends are all horrified, and the mass of Americans hate what’s happening; remember that. Donald is the problem, plus mass deception by Vladimir and Donald/MAGA. Stand with Americans themselves. They’re an amazing people, and they have saved our asses in the EU two or three times now. We owe them everything.
So bit of a change - something about Redshift!
I know right :-)
I’ve just been chatting to a fellow sufferer, he’d bumped into the new classic resize.
I’d heard about it some months back, but looking at what he found, we’ve seen some more detail, and so here I can describe the practical high-level upshot of what we’ve seen.
So, originally there was classic resize.
New cluster comes up, original goes read-only, all data copied over, redistributed, once all done, new cluster is up, and you swap to it. Note here billing is always for one cluster only. Problems : can perfectly well take days, no time estimate, not wholly reliable. Can freeze up.
Then elastic resize.
With this, number nodes changes, number of slices unchanged.
Elastic appears to resize quickly (30 mins say) but actually the new nodes have no data on local disk - it’s all in S3. Cluster then starts bringing data down from S3. So performance is smacked up until data finishes coming down to local disk. On an idle 8 node ra3.16xlarge, I saw about 1 TB coming down every 30 mins or so. If the cluster went back into normal use, that dropped to about 100 GB per hour.
Basically you need to wait after elastis for S3 to get down to local disk. Depending on your data size, can be a number of hours. There’s a view in RST to show how much data is on local disk and how much in S3.
So here we see something which is presenting itself as complete when it’s not, and the actual state of play is very hard to find in the system tables. I found it hard, and I know those damn tables.
Moving on, we now have new classic resize, and we now find more of the same - the resize comes back quickly and presents as “done” but in fact nothing has been done. All the work is now being done async, performance gets hit. Far as I know, not documented.
So what we saw was this; number of nodes changed, number of slices became correct, we were resizing down so we had two data slices to get rid of.
Looking at tables, we saw key dist had changed to even (there are not auto tables), and all the data from each extra slice had been downloaded en bloc and added to one slice, in the new cluster.
So before, we had a table with say 5m rows per slice. We lost two slices in total. Now we have on one node two slices with 10m rows each, all other slices and on all other nodes 5m rows. Where dist is even this works.
What you can see is a direct download from S3, no redistribution has occurred.
So now your table is borked. It’s not key, and it’s highly imbalanced. You can run a query on it, the data is there, but you wouldn’t want to, and if you had a real system which needs key for performance, hey, maybe you’re toast. Your system doesn’t actually meet business needs any more, because it’s too slow.
This is not documented, and you can’t stop it from happening, it takes days (as with the original classic resize) for RS to finish the background work of getting things back to key dist AND now you’re paying for this messed up cluster.
I’ve always said with RS don’t do resize. Never do it. RS is not resizable. First, it’s too slow (who can have a cluster out of action for days), second, there’s no time estimate, third, it’s not wholly reliable. It can freeze up.
Elastic resize is no good (compute slices, ra nodes running only specification number of slices at once so query time doubles, dc nodes run all slices at once so memory shortage problems), the new classic resize is worse than the original.
The way you solve this problem is the ETL system. Arrange ETL so you can run multiple clusters concurrently. When you want to resize, bring up a new cluster of the required size, populate it, move all users over, when no one is using the original, shut it down.
Do not resize Redshift clusters.
Home 3D Друк Blog Bring-Up Times Cross-Region Benchmarks Email Forums Mailing Lists Redshift Price Tracker Redshift Version Tracker Replacement System Tables Reserved Instances Marketplace Slack System Table Tracker The Known Universe White Papers