text stringlengths 9 408 |
|---|
[1323.62 --> 1327.58] Were you passing through a disk controller to that VM or what were you doing? |
[1328.12 --> 1338.48] So from some advice I got from, I'm sorry, I can't remember who it was on the self-hosted Discord, suggested using an LSI HPA. |
[1338.48 --> 1344.40] That HPA I then passed in through regular PCI pass through into TrueNAS. |
[1345.18 --> 1359.46] From there, regular mounted a very small two drive ZFS pool exposed via TrueNAS' built-in NFS sharing, which was super easy to do, which is great. |
[1359.46 --> 1363.06] Mounted that as normal through Etsy FSTab. |
[1363.06 --> 1371.64] And then from there, just pointing my Docker containers at it as if it was a directory in that container thing. |
[1371.72 --> 1379.14] And because it was running entirely over the internal network inside the machine, latency-wise, it was absolutely tiny. |
[1379.14 --> 1382.00] You're talking like fractions of milliseconds. |
[1382.48 --> 1385.44] But for some unknown reason, it didn't like it. |
[1385.48 --> 1392.94] Now, that might be because I went a little bit insane and decided to run TrueNAS scale rather than TrueNAS directly. |
[1393.36 --> 1399.58] So this may be a performance regression with TrueNAS scale because at the time I ran it, it was in alpha. |
[1399.58 --> 1402.88] And we know full well, don't run alpha software. |
[1403.28 --> 1405.62] But I did anyway, just to try it. |
[1406.32 --> 1410.22] Is scale the BSD one and core is the Linux one or is it the other way around? |
[1410.30 --> 1410.94] I can never remember. |
[1411.08 --> 1411.58] Other way around. |
[1411.78 --> 1416.36] So scale is the Linux-y one, which at the time was in alpha. |
[1416.72 --> 1417.70] So I tried it. |
[1417.90 --> 1420.40] It didn't work great. |
[1420.90 --> 1426.34] Yes, I have some thoughts on BSD this week after the whole WireGuard kernel merge thing. |
[1426.34 --> 1428.34] But let's not derail your... |
[1428.34 --> 1429.12] Yeah. |
[1429.58 --> 1431.54] That's not a fun discussion. |
[1431.76 --> 1432.38] It's all embarrassing. |
[1432.70 --> 1437.92] After I tried scale out, it didn't work too well for me. |
[1437.98 --> 1447.58] And I didn't want to try porting over to the actual core version of TrueNAS because keeping going with that process, |
[1447.58 --> 1455.96] it meant I had to deal with issues where that one storage VM, as it were, would have to stay up and alive sort of all the time. |
[1455.96 --> 1466.06] And if I needed to reboot that machine, that VM even, I would have to reboot sort of everything in the entire machine, which felt a bit like a faff. |
[1466.06 --> 1471.52] And eventually I concluded, well, if I want high performance, I want things to be easy. |
[1471.52 --> 1482.54] Then if I move the storage onto the actual host hypervisor under Proxmox, Proxmox has built-in support for ZFS because the OS is on ZFS. |
[1482.54 --> 1484.34] So it just works. |
[1484.34 --> 1485.96] It means latency is nice and low. |
[1485.96 --> 1496.88] And because the majority of my applications live in LXC, there's the nice benefit of rather than using NFS to mount all of my data, I can use bind mounts. |
[1497.22 --> 1498.98] And bind mounts are... |
[1498.98 --> 1507.78] You have no network latency because it's just passing directories into containers in much the same way that you give bind mounts to Docker containers. |
[1507.78 --> 1514.04] There's no extra latency, extra overhead, extra complexity around that. |
[1514.24 --> 1515.72] It just works. |
[1516.16 --> 1518.40] So it sounds like a lot of different application uses. |
[1518.40 --> 1523.76] And it sounds like you have a pretty distinct line between what is a container and what should be a VM. |
[1523.76 --> 1529.44] When I see complexity like this, I start to wonder, like, what's your backup strategy looking like? |
[1529.92 --> 1534.64] My backup strategy at the moment is on that one Docker LXC. |
[1534.84 --> 1544.40] There is a single instance of Duplicati, which backs up because that LXC specifically has access to my entire ZFS pool. |
[1545.12 --> 1550.76] I can just take the entire pool, compress it up with Duplicati and push it up to... |
[1550.76 --> 1559.68] Currently, there's a mixture of Backblaze and S3 for various bits, depending on how much I care about it. |
[1560.18 --> 1567.16] I intend at some point in the maybe not too distant future to switch parts of that out to Restic. |
[1567.60 --> 1569.42] Ah, good man. I was just about to ask you about that. |
[1570.40 --> 1572.80] Duplicati is such a pain in the ass sometimes. |
[1572.96 --> 1577.62] And, you know, it runs fine for weeks and weeks and then you log in and have a look at it. |
[1577.62 --> 1583.78] And in the intervening time, stuff's just broken for seemingly no reason. |
[1583.90 --> 1585.18] And it breaks silently. |
[1585.50 --> 1588.68] I can only configure it through the UI, which you know how much I love that. |
[1589.34 --> 1596.54] So Restic, I think, is something I'm really going to be looking into a lot in the future with Minio, with the S3 backend as well. |
[1596.76 --> 1598.78] So you can use Restic with that. |
[1598.96 --> 1601.78] So you use Backblaze and a little bit of S3. |
[1601.88 --> 1603.08] So is that Amazon S3? |
[1603.42 --> 1603.80] Yeah. |
[1603.80 --> 1605.04] Wow, made of money, boy. |
[1606.16 --> 1611.18] There's like a couple hundred meg in there just because I was testing stuff. |
[1611.66 --> 1615.12] Like this was before when I was only backing up like the tiny bits. |
[1615.26 --> 1619.80] And I concluded, well, if I'm just backing up the tiny bits, I might as well back up everything. |
[1620.26 --> 1621.90] Well, object storage is perfect for that. |
[1622.12 --> 1623.28] Exactly. 100%. |
[1623.28 --> 1626.04] Let me ask you this then, because I'm picturing the audience. |
[1626.16 --> 1628.24] Somebody's out there building their backup strategy right now. |
[1628.32 --> 1631.42] If you were going to start again today, start fresh, how would you build it better? |
[1631.52 --> 1632.66] What would you do differently? |
[1632.66 --> 1635.80] What message would you send to a past version of yourself? |
[1636.08 --> 1639.32] Past version would definitely be install backups sooner. |
[1639.88 --> 1646.62] I'm very glad that I've never had a case where I've had to actually restore from an offsite backup. |
[1646.62 --> 1658.74] I had one case where I had to restore from an offsite backup, which I still think is sort of a bug in Docker, where I was messing around with some weird Docker stuff. |
[1659.20 --> 1664.74] And I went to log into Nextcloud a few hours later and the entire Nextcloud bind mount was empty. |
[1665.44 --> 1666.42] Completely empty. |
[1666.52 --> 1668.36] No app data, no files, no nothing. |
[1668.50 --> 1669.32] Just gone. |
[1669.32 --> 1671.72] I have no idea why it happened. |
[1671.94 --> 1677.86] Fortunately, that morning I had done an onsite backup to a disconnected hard drive. |
[1678.20 --> 1686.72] So a few hours of syncing files back, I lost, as far as I'm aware, no real data because I'd not made any changes. |
[1686.84 --> 1688.50] But that was not fun. |
[1688.50 --> 1692.56] So definitely anyone listening, I think Alan Jude says it best. |
[1692.62 --> 1695.38] If it doesn't exist in three places, it doesn't exist at all. |
[1695.76 --> 1697.42] Oh, you had to go and reference Alan, huh? |
[1698.42 --> 1699.18] Yeah, I kind of did. |
[1699.24 --> 1704.04] But it's also it's such a good point is that now my data exists on my server. |
[1704.04 --> 1710.42] It also exists using ZFS send to a random shuck hard drive sitting on my desk. |
[1710.72 --> 1710.98] Good man. |
[1711.14 --> 1715.70] And then as well, like I say, backup to Backblaze, which has been absolutely great. |
[1715.78 --> 1720.98] I did for a small amount of time decide, hey, what if I roll my own Backblaze? |
[1721.84 --> 1722.86] Would not recommend it. |
[1723.08 --> 1731.06] I had a $10 machine from Kim Sufi, which are a very cheap, dedicated server host out of France. |
[1731.06 --> 1735.20] They're a subsidiary of OVH and we all know how OVH are doing right now. |
[1735.44 --> 1737.66] Yeah, I hear their business is on fire at the minute. |
[1737.90 --> 1738.54] Oh, Alex. |
[1738.80 --> 1739.34] Too soon. |
[1739.50 --> 1740.04] Too soon. |
[1740.30 --> 1740.88] Too soon. |
[1741.40 --> 1742.74] I thought it was funny. |
[1742.92 --> 1743.08] Yeah. |
[1743.12 --> 1747.04] So I tried running that and Minio and it worked fine. |
[1747.32 --> 1753.98] But once I realized that I was backing up a server I owned to a server I owned, it was great. |
[1754.28 --> 1757.04] But it was the Kim Sufi box was a bit finicky. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.