2025-01-12 GnuCash IRC logs

02:37:01 *** sunyibo has quit IRC
02:59:47 *** brook_ has joined #gnucash
03:00:44 *** brook has quit IRC
06:56:07 *** jonakeys has quit IRC
07:15:08 *** jonakeys has joined #gnucash
10:12:09 *** aj has quit IRC
10:13:51 *** aj has joined #gnucash
10:13:51 *** ChanServ sets mode: +v aj
11:22:59 *** brook has joined #gnucash
11:24:04 *** brook_ has quit IRC
12:55:49 *** jralls-m1 is now known as jralls
12:56:19 *** ChanServ sets mode: +o jralls
13:01:50 <jralls> CDB-Man, can you elaborate on adding another currency field to transactions? What currency would you store and how would you use it?
13:03:50 <jralls> warlord, have you given any thought to upgrading the Windows build VM to Win11 before Win10's end of support?
13:04:23 <warlord> jralls, Honestly, haven't thought about it at all.
13:05:00 <warlord> I would need to first acquire a Win11 install media and license... Also, I'm not 100% sure it will run in an ovirt VM
13:07:16 <jralls> It would be surprising of it didn't, it's been out for 3 years now. That's plenty of time for Ovirt to have done whatever they needed to. The main issue is whether code's hardware is new enough.
13:10:06 <warlord> jralls, the issue is more that I'm running an older version of oVirt, version 4.3.10, which does not support it.
13:10:46 <warlord> ... and upgrading ovirt requires updating the whole OS on the system, because the host is running EL7
13:12:03 <jralls> And that's too hard?
13:12:27 <warlord> Yes. I've got only the single host.
13:12:51 <warlord> Also, I feel like the ovirt team has deprecated the hosted-engine (which is what I'm using)
13:13:19 <warlord> ... and I'm frankly scared to try to do a full system upgrade, for fear if it breaks I won't be able to back it out.
13:18:04 <jralls> Well, RHEL7 goes out of support on June 30, before even Win10 does on October 14. It being somewhat insane to run a public server on an out-of-support OS I think you're going to have to conquer your fear. Is the system on its own partition so if the worst happens you can just nuke it and do a clean install?
13:20:35 <warlord> I'll point out that this is one of the reasons I wanted a 2nd machine... because it would allow migrations and updates without downtime. (Technically would need 3 machines)
13:21:54 <warlord> Yes, the OS is on its own LVM LV.
13:24:27 <jralls> We can tolerate a week's downtime. It wouldn't be the first time. But maybe we need a new server anyway if the CPU isn't up to Win11's requirements.
13:25:31 <warlord> The CPU should be fine, it's that ovirt 4.3 does not emulate TPM2.0.
13:28:25 <warlord> I am fairly sure I could do it if I could safely upgrade the system, but IMHO the best approach would be to "reinstall" ovirt into a new cluster and migrate the existing VMs -- and then re-incorporate the existing hardware into the "new" cluster.
13:29:22 <warlord> I bet HW prices have come down -- although I haven't spec'ed out new systems recently.
13:29:59 <warlord> Biggest issue would be the storage system -- do I use GlusterFS (which appears to have been abandoned?) or do I figure out some other shared storage system?
13:34:03 *** jralls has quit IRC
13:34:12 *** jralls has joined #gnucash
13:34:13 *** ChanServ sets mode: +o jralls
13:35:43 <jralls> PC Health app says the CPU isn't fine. It says it's an "Intel Core (Broadwell)". Is Ovirt reporting the wrong thing?
13:36:25 <warlord> model name : Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
13:41:07 <warlord> (that's what /proc/cpuinfo reports from the host)
13:41:16 <jralls> Not on the list: https://learn.microsoft.com/en-us/windows-hardware/design/minimum/supported/windows-11-22h2-supported-intel-processors.
13:45:42 <warlord> So, clearly there are several issues with installing win11 VM.
13:45:44 <jralls> An per https://en.wikipedia.org/wiki/List_of_Intel_Xeon_processors E5-2620 V4 is indeed Broadwell.
13:47:14 <warlord> If you thnk about it, the current hardware is AT LEAST 7 years old..
13:47:34 <warlord> My setup notes says:
13:47:35 <warlord> # 2016-10-22 -- ovirt-0.ihtfp.org -- Install CentOS 7.2
13:48:44 <warlord> So 8 1/2 years, if my math is right?
13:49:31 <jralls> That sounds about right. Isn't that about when you moved into your house?
13:50:43 <jralls> So it's time--maybe past time--to raid the cookie jar and get a new server.
13:54:45 <jralls> https://lists.gnucash.org/mailman/private/gnucash-core/2016-October/000658.html matches your log. ;-)
13:54:50 <warlord> I moved into my new house Dec 2019
13:55:07 <warlord> (bought it in Oct 2017)
13:55:30 <warlord> Yes, it might be about time to "upgrade". I just emailed the current tip-jar balance.
13:57:15 <jralls> BTW, you say in that proposal that the previous machine was "7, almost 8 year[s] old".
13:58:29 <warlord> Yeah, at THAT time the old machine was probably that age!
13:58:50 <warlord> The current machine still runs fine; I think we can have a mixed-machine ovirt cluster.
14:00:23 <warlord> Or, apparently not -- a cluster needs to share a CPU-type.
14:09:28 <jralls> As for GlusterFS, their Github repo is still getting commits--the latest one was December 3--but it seems to be only a few in the last year. Can't be exact, going to the history page causes a rate-limit error.
14:13:04 <jralls> https://en.wikipedia.org/wiki/Comparison_of_distributed_file_systems lists about 20 alternatives.
14:17:38 <warlord> Looks like Gluster of Ceph are options.
14:18:13 <jralls> Gluster of Ceph seems redundant?
14:18:24 <warlord> Sorry, s/of/or
14:22:21 <jralls> Well, you've already got Gluster and know how to use it. On the other hand Ceph is vastly more active, over 150,000 commits with 30 in the last 4 days.
14:23:01 <warlord> I am not using Gluster currently.
14:23:16 <warlord> I'm actually using NFS.. But that's only because I've just got the single server.
14:25:30 <jralls> Oh, OK>
14:25:30 <warlord> Ideally I'd love to have a 3-system cluster, which would allow me to take one system down for maintenance without losing quorum.
14:26:02 <warlord> I wonder if I could upgrade the current system hardware just by replacing the CPUs? But I don't know if the mobo would support modern CPUs.
14:28:09 <jralls> Probably not. Intel seems never to keep the same socket across generations. There's usually also chipset compatibility issues. But you could, at least in theory, just swap MoBos as long as the current enclosure can supply enough power and get rid of the heat.
14:29:30 <warlord> Fair... I wonder if I should spec out a 3-node cluster?
14:30:49 <jralls> That wouldn't fit in the enclosure. ;-)
14:31:43 <warlord> LOL. No, it wouldn't.
14:33:06 <warlord> the 3 nodes would include 3 enclosures.
14:33:32 <warlord> -- i would need to keep the current system up and running to migrate to the new one
14:33:39 <warlord> (I think)
14:37:07 <jralls> It would be nice, but like I said we can tolerate a week of downtime. Have you built a cluster before?
14:39:03 <warlord> No, I've always had single-machine systems. But the docs are pretty straight-forward.
14:39:24 <warlord> FWIW, when I spec'ed out a machine in 2021 it was about $5k
14:41:39 <warlord> Of course, I would need to go re-spec everything with "modern" H/W.
14:41:43 <jralls> > "But the docs are pretty straight-forward." < Famous last words. ;-). So add a week or two..
14:42:42 <warlord> LOL... The benefits of "starting from scratch" is that it would allow a parallel development, so I wouldn't have to take my infra offline to set up the new cluster.
14:43:59 <jralls> And consider the balance of ssd vs spinning rust.
14:45:01 <jralls> Right. The extra complexity rules out a MoBo swap solution.
14:46:43 <warlord> Are you suggesting going with more SSD? or more spinning rust?
14:47:10 <warlord> Of course, so far I'm 0/2 with previously-spec'd parts being available today! LOL.
14:47:32 <warlord> It's going to take a longer time to re-spec than I thought. Certainly not a 5-minute task.
14:52:09 <jralls> No, a lot changes in 4 years. As for SSD vs. rust, I don't know what the storage usage is so I don't know whether the current 2-3 split is right, nor if 5TB is enough. I guess that since you got 2 of each they're mirrored?
14:53:19 <warlord> Yes, everything is mirrored (both spinning rust and SSD)
14:54:22 <warlord> If I go to Ceph, though, I'm not sure if I necessarily need to mirror each SSD locally? I would think I would use a part of spinning rust for local OS (can be smaller disks) -- and then use a single or double SSD, non-mirrored, for Ceph.. And let Ceph do the raid across a 10G network.
14:55:28 <jralls> No idea about Ceph, I've never done distributed FS or clusters.
15:01:06 <warlord> honestly, neither have I! LOL
15:09:24 <jralls> Right, two more weeks! Do you really want to spend all that time on this? Even if it blows up a MoBo swap and OS upgrade/replacement will only take a couple of days.
15:22:14 <warlord> Honestly, yes. longer-term I'd rather have the redundancy of a cluster in place.. It will give me the flexibility to more-easily update the infra in the future.
15:23:11 <warlord> The downside is that things like 2U chassis cases are almost twice the cost they were in 2021.. A supermicro 2U 8-bay(hotswap) case is $868 (+tax&s/h). In 2021 it was $465.
15:29:44 <jralls> Yeah, inflation. It'll be a lot more in a couple of weeks if Trump follows through on his tariffs.
15:30:33 <warlord> Yeah... So clearly should order now.
15:37:06 <warlord> Honestly, I don't even know the "order" of the XX-Lake systems to know what to look for!
15:44:45 <jralls> The current generation is Raptor Lake. But consider AMD, they're widely considered to be better.
15:45:34 <jralls> But don't sweat it too hard, anything with a reasonable core count will be fine.
15:51:37 <CDB-Man> [2025.01.12 13:01:50] <jralls> CDB-Man, can you elaborate on adding another currency field to transactions? What currency would you store and how would you use it?
15:52:16 <CDB-Man> https://i.imgur.com/63ZiBHP.png
15:53:14 <CDB-Man> err, rather this one: https://i.imgur.com/S7sHgUB.png
15:53:21 <CDB-Man> correction on exmaple #2
15:54:09 <jralls> That doesn't make sense to me. Please separate transaction and splits.
15:55:07 <CDB-Man> there are 8 transactions here each with 2 splits -- I havew not yet inserted the trading splits (and I also found another copy paste error in the example)
15:55:16 <CDB-Man> 7 transactions*
15:57:25 <CDB-Man> https://i.imgur.com/Awro9nT.png ** with the copy/paste fixed -- each one of these are 7 standalone transactions, with 2 splits for what would be recorded in a commercial solution
15:58:41 <CDB-Man> hmm, I think I need to re-do the examples where I am spending non-CAD -- built this too quickly
15:59:07 <CDB-Man> no, they should be fine, I'm spending the correct foreign currency amount, and everything equiates to $100 CAD
15:59:19 <jralls> It still doesn't make sense. You said Friday night that you wanted to add a currency field to the transaction, but it looks like you're adding a field to each split instead.
16:00:25 <CDB-Man> hmm, that may have been poor use of terminology on my part last Friday then
16:01:12 <CDB-Man> in my mind, you would need it each split, as that is how you would handle multiple (ie 3+) currencies in a single transaction
16:02:07 <CDB-Man> you will see that the transaction currency is CAD in all 7 examples though, that's why they are all $100 CAD
16:03:06 <warlord> jralls, I'm currently looking at "xeon scalable".. But man, prices are craycray compared to what I was seeing in 2021.
16:04:41 <jralls> CDB-Man, That's another terminology fail then: All of the amounts are CAD, but amounts are in the split-account's commodity. Values are in the transaction currency.
16:06:00 <CDB-Man> ... right, "amount" and "value" are not synonyms in gnucash
16:06:03 <jralls> CDB-Man, in your examples there are two "Currency" fields that are the same (USD or HKD) in each one. It all looks redundant.
16:06:59 <CDB-Man> 2 curency fields are only required if you had to do all conversions at the transaction and split level -- if priceDB can confirm a USD quoted stock directly to CAD, you could eliminate 1 set of currency
16:07:24 <CDB-Man> priceDB can convert**
16:08:28 <CDB-Man> the middle 3 columns, "foreign amiunt" / FX / currency could be eliminated if we can for example, generate the HKD stock's value directly into CAD via a double conversion in priceDB
16:08:30 <jralls> PriceDB shouldn't be involved in recording transactions. It gets a copy of the transaction price to make sure that it always has a price, but transactions have to stand on their own.
16:09:12 <CDB-Man> sure -- my point is, if there is a way we can convert an amount of HKD stock commodity directly in CAD, and feed that into the value of the split, then you could eliminate a set of conversions
16:10:26 <jralls> If there's a split that has BBB->HKD and one that has HKD->CAD then we can calculate BBB->CAD without recording any extra data. We just need to change the UI to collect both rates and do so only once per transaction instead of once per split.
16:10:40 <CDB-Man> the ultimate goal is to always return, in this case, a CAD value for each split, since transaction currency = book corrency = CAD in this example. my screenshot here is just illustrating the 2 step conversion:
16:10:40 <CDB-Man> 1) BBB stock into HKD amount
16:10:40 <CDB-Man> 2) HKD amount into CAD value
16:11:53 <CDB-Man> indeed, what you desribed works too -- my screenshot here is verbose on purpose to "show the homework" -- if you can physically achieve it in fewer steps, by all means
16:12:15 <jralls> warlord, that would be money down the drain. Those processors aren't anything like what we need.
16:14:14 <warlord> I'm so out of the loop, I honestly have no idea what it is we actually need.
16:17:21 <jralls> warlord, a plain CPU with plenty of cores for running multiple VMs without too much context-swapping. We don't need any vector processors with silly names like Analytics Engine, nor do we need a CPU with silicon devoted to talking to these vector processors.
16:18:24 <warlord> Fair point. Is there a "find the right CPU for you?" configurator somewhere?
16:19:13 <jralls> CDB-Man, on a somewhat related note, here's a different approach entirely: https://bugs.gnucash.org/show_bug.cgi?id=799041. Mr. Bruniges says that he needs his book to balance in each of 3 currencies. I think he's mostly unique, but I'm interested in whether that's compatible with IAS.
16:20:37 <jralls> warlord, I don't know of one. But like I said it's not worth a lot of effort. Get something with 12 or 16 cores @ 3GHz and it'll be fine.
16:21:47 <CDB-Man> that's quite the lengthy thread -- is there 1 of those attechments that has the final of what he's trying to achieve? I see mention of the attachments being edited and such
16:24:37 <CDB-Man> on his matter of fractions, he could simply define eall of his securites to use the maximum number of zeros
16:26:18 <CDB-Man> I don't know what he means by "multi-curency book" -- in the IAS world, the master chart of accounts must be in a single functional currency -- perhaps he is either using "multiple charts of accounts" that have a back-end consolidations to a single chart of accounts (and this consolidation is transparent to the user)
16:26:58 <CDB-Man> or perhaps it is already a single chart of accounts, but the UI of his software natively "displays" anty foreign currency accounts with a UI that automatically converts everything to said currency
16:28:33 <CDB-Man> that or, when he says "multiple currency", he actually is just referring to the fact that he deals in multiple currencies, but ultimately everything still resolves to a single currency because it has to do so
16:31:52 *** afdee1c has joined #gnucash
16:39:40 <warlord> jralls, what about the Xeon Silver 4310? Ice Lake 2.1GHz?
17:04:08 <jralls> warlord, it should be OK, it's on the list. But an i7-14700 is faster than 2 of those and cheaper than one of them, albeit with the same number of cores as 1 Xeon.
17:06:40 <warlord> By "faster than 2 of those" do you mean "twice as fast"?
17:07:15 <jralls> https://www.cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+9+5900X&id=3870 vs https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+Silver+4310+%40+2.10GHz&id=4322&cpuCount=2.
17:07:56 <jralls> So the i7 is faster than 2 Xeons even though the 2 Xeons combined have twice as many cores.
17:08:41 <jralls> Oops, the first link should have been https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-14700&id=5852.
17:09:22 <jralls> On single-thread performance the i7 is almost 3x faster than the Xeon.
17:12:49 *** fell has joined #gnucash
17:12:49 *** ChanServ sets mode: +o fell
17:13:54 <jralls> That first link is for an AMD comparable to the i7 but without the Raptor Lake instability issue https://en.wikipedia.org/wiki/Raptor_Lake#Instability_and_degradation_issue.
17:15:49 *** afdee1c has quit IRC
17:16:29 <jralls> CDB-Man, I think Bruniges means that he deals in several currencies. In one post he says he has a simple script that adds balancing amounts @ 0 value splits to make sure that his books balance in all currencies.
17:17:39 <CDB-Man> I wonder if his balancing amounts are to deal with the fractional roundsing, or if he means something else
17:19:43 <CDB-Man> there's no *requirement per se* in IAS to balance your foreign currency amounts. there's also a concept in IAS of froeign "cumulative translation adjusemnt" gain/loss, which I'm by no means the export (it's fairly complex), but it's the IAS was to record the unrealized gain/loss on currency relative to your home currency
17:19:50 <CDB-Man> https://ifrscommunity.com/knowledge-base/ias-21-effects-of-changes-in-foreign-exchange-rates/
17:20:02 <jralls> Probably rounding. He implies in that last post that he's already making sure that the transaction currency is his book currency.
17:24:18 <warlord> jralls, Interesting. Single-thread performance is important in terms of how fast a VM can run operations. So the 3x single-thread speed is probably important on a per-VM basis. On the other hand, the 192GB RAM limit might be a problem to support many VMs on the system.
17:28:22 <jralls> warlord, that would be a lot of VMs. Besides, how many 3-year-old MoBos can support even that much memory?
17:29:28 <warlord> I've got 256GB in the current server
17:30:17 <warlord> If I went with the Xeon, I'd spec out 2x CPUs per board and 4x 64GB RDIMMS
17:34:17 <jralls> How many VMs are you running, and what the heck are they doing to need that much RAM? Or, flipping it, how much of that memory is actually getting used?
17:36:03 <Simon> if you're doing CI, try to build in RAM
17:36:25 <Simon> it makes everything so much faster and avoids doing loads of disk I/O that'll get discarded
17:36:26 <jralls> Simon, We're not. CI runs on Github.
17:37:56 <jralls> Aside from that a 1GB ramdisk would be plenty for building GnuCash and its dependencies.
17:40:30 <warlord> ovirt reports 125.8G used.
17:43:59 <warlord> So perhaps I don't need 256G/server -- I can get away with 128G ea, so that 2x can support what the current server runs.
17:45:09 <warlord> Granted, the RAM is only $175/64G-DIMM so removing that didn't significantly affect total cost.
17:48:09 <jralls> Right, the reason for the discussion is i7's 196GB limit, which on most MoBos would have to be 6x32.
17:50:41 <jralls> I suspect ovirt is reporting how much memory you've allocated to running VMs rather than how much is actually being used right now. Nothing's building so they're just passing mail around and serving a couple of webpages. That doesn't use 126G of RAM.
17:51:07 <warlord> Also, it looks like you can't get a dual-i7 mobo.. But again, I suppose if I am expecting to split the load across 2 servers, it should be fine having a single CPI
17:51:33 <warlord> There are 30 running VMs
17:51:47 <warlord> (it's not all gnucash load)
17:52:18 <jralls> Right, that's the difference between Xeons and Core CPUs. The Core ones are single-CPU-per-board.
17:52:53 <jralls> "It's not all gnucash load" < Then why should GnuCash pay for it?
17:55:35 <jralls> CDB-Man Interesting article, thanks. Makes me glad I don't do foreign currency!
17:58:19 <warlord> That's fair question.. I have no personal need to upgrade the hardware at this time, and none of the existing load has that need, either. The only input for an upgrade is the gnucash request for Win11, which apparently wont work on the current platform.
17:58:56 <jralls> Which GnuCash also paid for.
17:59:07 <warlord> GnuCash paid for some of it, not all of it.
18:05:08 <jralls> Well, GnuCash's need is for about $500 worth of computer to do nightly builds and it looks like you're asking for GnuCash to fund ~$15,000 worth of upgrades to get there.
18:05:38 <jralls> The obvious alternative is to move the nightly builds to Github actions.
18:06:44 <warlord> Can GA do that?
18:07:49 <jralls> Yup. CI is 95% of the way there, we just need to add the packaging.
18:14:22 <warlord> okay.
23:10:01 *** jonakeys has quit IRC
23:10:08 *** jonakeys has joined #gnucash