RK3566之armvirt转发能力

2024-03-02 作者： Fisherworks

一言以蔽之，同用双核，略逊Pi 4B，明显不如RK3399。虽然制程和频率双双强过BCM2711（Pi 4B），但A55相比A72应该还是有明显差距的…吧，当然发行版对比也略有差异（尽管都基于debian），无法精确对比哈。

至于想劝俺上什么iStoreOS就算了，原因在之前提过，尽管估计物理跑NAT（开启软件流量分载）大概率能接近跑满千兆转发，但选型喜好是很个人的事情，总觉得所谓all-in-one在架构上反过来（upside down）不好玩。

数据上，iperf3上行700左右，下行850，看起来即使另外两核跑个容器带点儿负载，保障个500Mbps宽带应该没啥鸭梨。

innobackupex吃光root分区

2024-02-29 作者： Fisherworks

Take a look at this if your innobackupex insists on eating up most (or all) of your root partition, as playing hide and seek with you especially.

近两天总收到奇怪的推送，磁盘占用超90%，没采取人为任何措施，又自动降至正常……就很离谱。去翻了一下云监控的磁盘使用率，发现居然是系统盘，而非数据盘（后者占用率不高，多日无明显变化），就更奇怪了。

前者么，不逮住问题发生时的话，根本无从查起；考虑其发生规律，基本跟mysql冷备的时间周期比较吻合。

所以先去看mysql冷备文件的状态，就开始出问题了（截图欠奉）。

每个备份文件的结束时间都很早，从以往常见的上午9~10点提前到了凌晨4~5点。
且文件的大小也不太对劲，降了一半不止（在未删除数据的前提下）。

好办，备份文件会定期rotate但log不会，不删就会一直存，看log就好。

一看不要紧，近两天备份都是失败告终。

..........
240209 05:05:01 >> log scanned up to (58627078203151)
240209 05:05:03 >> log scanned up to (58627086878169)
innobackupex: Error writing file '/tmp/xbtempPr6mT8' (Errcode: 28 - No space left on device)
xtrabackup: Error: write to logfile failed
xtrabackup: Error: xtrabackup_copy_logfile() failed.

1

2

3

4

5

6

..........

240209 05:05:01 >> log scanned up to (58627078203151)

240209 05:05:03 >> log scanned up to (58627086878169)

innobackupex: Error writing file '/tmp/xbtempPr6mT8' (Errcode: 28 - No space left on device)

xtrabackup: Error: write to logfile failed

xtrabackup: Error: xtrabackup_copy_logfile() failed.

这里的 /tmp 就很迷，我印象中所有临时文件都不应该去系统默认的temp目录，索性检查一下。

完整阅读本篇»

从lxd迁移到incus

2024-02-27 作者： Fisherworks

Let’s talk about migration from lxd to incus. As for the reason, well I guess we have seen too much of this kinda “show” in the past half decade (just like CentOS to Rocky) so here is what happened in case you’re interested.

Technically I won’t take this as a tutorial since with the well-prepared tool lxd-to-incus by Stéphane Graber (Author of both lxd and incus) there’s nothing to do but simply running it, then everything seems working well.

PS: Tested PASS on both Debian Bookworm AMD64 (on Linux KVM based vm) and Armbian Bookworm ARM64 (on RK3566 based SBC).

这些年，同类剧本真是没少看。一言不合就闭源，一言不合就拉分支，完全不新鲜，具体就不多说了，开工干活。其实吧，作者给了迁移工具 lxd-to-incus ，所以实话说没啥可干的，执行一下就可以验证迁移后的结果了。

先检查系统现状。Check the system status at first.

# root @ bookworm-vm-host in ~ [17:55:15]
$ uname -a
Linux bookworm-vm-host 6.1.0-18-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.76-1 (2024-02-01) x86_64 GNU/Linux

# root @ bookworm-vm-host in ~ [17:55:20]
$ cat /etc/os-release
PRETTY_NAME="Debian GNU/Linux 12 (bookworm)"
NAME="Debian GNU/Linux"
VERSION_ID="12"
VERSION="12 (bookworm)"
VERSION_CODENAME=bookworm
ID=debian
HOME_URL="https://www.debian.org/"
SUPPORT_URL="https://www.debian.org/support"
BUG_REPORT_URL="https://bugs.debian.org/"

# root @ bookworm-vm-host in ~ [17:55:28]
$ lxc --version
5.0.2

# root @ bookworm-vm-host in ~ [17:55:35]
$ lxc list -c nsum4N
+-------------+---------+-----------+--------------+------------------------+-----------+
|    NAME     |  STATE  | CPU USAGE | MEMORY USAGE |          IPV4          | PROCESSES |
+-------------+---------+-----------+--------------+------------------------+-----------+
| d10-test    | RUNNING | 5s        | 94.54MiB     | 192.168.143.138 (eth0) | 6         |
+-------------+---------+-----------+--------------+------------------------+-----------+
| d12-ngx     | RUNNING | 6s        | 107.81MiB    | 192.168.143.2 (eth0)   | 19        |
+-------------+---------+-----------+--------------+------------------------+-----------+
| r8-app1     | RUNNING | 8s        | 213.93MiB    | 192.168.143.129 (eth0) | 235       |
+-------------+---------+-----------+--------------+------------------------+-----------+
| u18-redmine | RUNNING | 19s       | 402.67MiB    | 192.168.143.158 (eth0) | 16        |
+-------------+---------+-----------+--------------+------------------------+-----------+

# root @ bookworm-vm-host in ~ [17:55:39]
$

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

# root @ bookworm-vm-host in ~ [17:55:15]

$ uname -a

Linux bookworm-vm-host 6.1.0-18-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.76-1 (2024-02-01) x86_64 GNU/Linux

# root @ bookworm-vm-host in ~ [17:55:20]

$ cat /etc/os-release

PRETTY_NAME="Debian GNU/Linux 12 (bookworm)"

NAME="Debian GNU/Linux"

VERSION_ID="12"

VERSION="12 (bookworm)"

VERSION_CODENAME=bookworm

ID=debian

HOME_URL="https://www.debian.org/"

SUPPORT_URL="https://www.debian.org/support"

BUG_REPORT_URL="https://bugs.debian.org/"

# root @ bookworm-vm-host in ~ [17:55:28]

$ lxc --version

5.0.2

# root @ bookworm-vm-host in ~ [17:55:35]

$ lxc list -c nsum4N

+-------------+---------+-----------+--------------+------------------------+-----------+

+-------------+---------+-----------+--------------+------------------------+-----------+

| d10-test | RUNNING | 5s | 94.54MiB | 192.168.143.138 (eth0) | 6 |

+-------------+---------+-----------+--------------+------------------------+-----------+

| d12-ngx | RUNNING | 6s | 107.81MiB | 192.168.143.2 (eth0) | 19 |

+-------------+---------+-----------+--------------+------------------------+-----------+

| r8-app1 | RUNNING | 8s | 213.93MiB | 192.168.143.129 (eth0) | 235 |

+-------------+---------+-----------+--------------+------------------------+-----------+

| u18-redmine | RUNNING | 19s | 402.67MiB | 192.168.143.158 (eth0) | 16 |

+-------------+---------+-----------+--------------+------------------------+-----------+

# root @ bookworm-vm-host in ~ [17:55:39]

$

然后根据文档，安装incus。Now install the incus (by Zabbly) according to the documentation.

从安装过程看，至少创建了一个用户，两个组，以及不少于6个systemd服务。

完整阅读本篇»

Rasp Pi 5B vs. Intel N100?

2024-01-15 作者： Fisherworks

我先给一句话总结，之后再慢慢解释：

当两个不同的细分市场的产品，在各自的领域站稳脚跟（也即拿到主要份额）后，无论出于产品代次的需要、抵御外部威胁的需要、抑或资本市场的需要，均会体现出向另外一个领域渗透的特征。

Talking about a specific industry, when different products designed for different market segments obtained dominant shares of their own segments, almost every of them would consider invading neighbors (segments around) – no matter the needs come from product roadmap, external threats or even the capital market.

解释：一个字，卷。

最近有人问我，自制软路由/NAS是买arm还是x86_64。如果是小10年（Pi 2B上市）甚至5年前，我只会说，萝卜青菜各有所爱，不可一概而论，还得看需求更关注什么。

前者具备体积、功耗/散热、价格、专芯专用（如hw-nat）的优势，后者则表现出强性能、AIO的独特能力。

而放眼如今，早已是 –