text stringlengths 7 369 |
|---|
[2048.24 --> 2050.82] gets given PID1, process ID1. |
[2051.54 --> 2055.26] And when that PID dies, the container dies with it. |
[2055.26 --> 2057.38] The same, by the way, is true of your Linux server |
[2057.38 --> 2060.44] or your Linux desktop or whatever you're running Linux on. |
[2060.80 --> 2062.98] When PID1 dies, the system dies, |
[2063.06 --> 2065.66] and typically that PID1 is systemd these days. |
[2066.44 --> 2072.10] In an LXC, what happens is you get an init system like systemd. |
[2072.94 --> 2075.92] My biggest gripe with LXC generally, |
[2076.12 --> 2077.16] and there's two really, |
[2077.16 --> 2080.12] one is that it's an extra layer of complexity |
[2080.12 --> 2082.04] that not that many people are using. |
[2082.18 --> 2085.00] Some people in our self-hosted community are using it, |
[2085.22 --> 2087.22] but generally it's on the fringes |
[2087.22 --> 2088.42] compared to the number of people |
[2088.42 --> 2089.90] that are using Docker containers. |
[2090.56 --> 2093.78] The second thing is that running more than one process |
[2093.78 --> 2095.90] in an isolated environment like that |
[2095.90 --> 2098.36] is a bit of an anti-pattern. |
[2098.50 --> 2099.52] And the reason I say that |
[2099.52 --> 2101.46] is because containers are designed |
[2101.46 --> 2104.38] to be this black box almost |
[2104.38 --> 2107.30] of just a single block with a single purpose. |
[2108.04 --> 2109.10] And if you start trying to run |
[2109.10 --> 2111.30] more than one service inside that container, |
[2111.56 --> 2113.52] so let's say systemd plus a web server, |
[2114.04 --> 2115.52] on the face of it, |
[2115.58 --> 2117.78] you're only running one service, right? |
[2117.82 --> 2118.42] A web server. |
[2118.62 --> 2119.70] But actually you're running two, |
[2119.82 --> 2122.22] you're running an init system plus a web server, |
[2122.38 --> 2123.64] which is a whole bunch of overhead |
[2123.64 --> 2124.72] that you A, don't need, |
[2125.10 --> 2127.06] and B, technically speaking, |
[2127.20 --> 2128.84] is, like I said, a bit of an anti-pattern. |
[2128.84 --> 2132.40] And so the other thing it gives you is some issues, |
[2132.54 --> 2134.06] well not issues, it's not the right word, |
[2134.14 --> 2137.66] but added complexity around passing hardware through. |
[2137.80 --> 2139.44] So like QuickSync, for example, |
[2139.52 --> 2141.38] which I think is what we talked about earlier in the show. |
[2141.94 --> 2144.92] If you're running QuickSync in a Docker container, |
[2145.10 --> 2148.66] it's a couple of lines to pass through the render device. |
[2149.44 --> 2151.78] To QuickSync, running in Plex or whatever, |
[2152.94 --> 2153.76] it's no big deal. |
[2154.08 --> 2156.46] With LXC, there can be some compatibility issues. |
[2156.80 --> 2157.62] It's a similar thing, |
[2157.62 --> 2158.72] you've got to pass it through, |
[2159.32 --> 2160.06] but there, |
[2160.26 --> 2162.96] I can never remember if it has a distinct lock |
[2162.96 --> 2166.82] on the GPU once you pass it through to an LXC or not, |
[2166.86 --> 2168.72] so somebody in the chat could update in real time, |
[2168.80 --> 2169.78] I'd appreciate it. |
[2170.34 --> 2172.06] But then the other thing you've got to consider is, |
[2172.16 --> 2173.80] if you want to run multiple services |
[2173.80 --> 2176.04] in one isolated environment, |
[2176.90 --> 2179.58] that's exactly what a VM has been designed to do. |
[2180.22 --> 2183.42] The tooling around VMs is extremely mature at this point. |
[2183.42 --> 2184.62] They're very, |
[2184.78 --> 2185.72] very well understood. |
[2186.88 --> 2188.60] And I don't know, |
[2188.76 --> 2193.12] LXCs for me fall into this awkward middle ground |
[2193.12 --> 2196.04] of not quite knowing who they're for. |
[2196.54 --> 2198.20] It's not to say they're a bad technology, |
[2198.20 --> 2201.86] because I think LXCs do have their place. |
[2202.56 --> 2202.90] Just, |
[2203.86 --> 2204.52] they just don't, |
[2204.88 --> 2206.14] in my world personally, |
[2206.58 --> 2206.88] you know? |
[2207.60 --> 2207.78] Yeah, |
[2207.86 --> 2208.00] well, |
[2208.08 --> 2208.18] and, |
[2208.24 --> 2208.66] you know, |
[2208.70 --> 2210.18] I think they're also a little more popular |
[2210.18 --> 2212.42] in the Ubuntu ecosystem as well, |
[2212.52 --> 2214.80] which is probably just not as prevalent |
[2214.80 --> 2216.00] in the North America area. |
[2216.26 --> 2216.48] So, |
[2216.62 --> 2218.20] I think there's that element as well. |
[2218.56 --> 2219.58] I like what you're saying, |
[2219.62 --> 2219.70] though, |
[2219.70 --> 2220.22] about VMs, |
[2220.22 --> 2222.00] just to step back a little bit there. |
[2222.24 --> 2224.02] Your point about the VM |
[2224.02 --> 2225.26] is sometimes the right tooling, |
[2225.34 --> 2227.08] people shouldn't feel bad about it, |
[2227.10 --> 2228.92] I think is underscored by the fact |
[2228.92 --> 2231.14] that VM performance, |
[2231.38 --> 2232.14] it's like, |
[2232.30 --> 2234.44] it's not your grandpa's VM anymore. |
[2234.44 --> 2235.86] On modern hardware, |
[2236.06 --> 2237.76] with modern virtualization software, |
[2238.16 --> 2240.06] VMs can be extremely performant, |
[2240.14 --> 2241.90] way better than they were |
[2241.90 --> 2242.74] when we were emulating |
[2242.74 --> 2243.90] the entire hardware stack. |
[2243.98 --> 2244.66] Because these days, |
[2245.02 --> 2246.02] not only do we have |
[2246.02 --> 2247.42] para-virtualized devices, |
[2247.72 --> 2248.88] but there's also just been |
[2248.88 --> 2250.00] a lot of plumbing in the kernel |
[2250.00 --> 2252.62] to get virtual driver requests |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.