Skip to content

seaworld008/centos7-lighthouse / skills/centos7-almalinux-upgrade

centos7-almalinux-upgrade

Flagged

把老旧 CentOS 7 / CentOS 8(及 RHEL7/8 系)服务器原地升级到最新 AlmaLinux 10(经 ELevate/Leapp 7→8→9→10,CentOS 8 先用 almalinux-deploy 转换),并保证业务、容器、nginx、kubectl 等升级后照常工作。凡是用户提到 CentOS 7 / CentOS 停服 / EOL / 老系统升级 / 迁移到 AlmaLinux / Rocky / CentOS 继任者 / ELevate / leapp / 系统漏洞要升级系统、批量升级阿里云/华为云/AWS 上的 CentOS 服务器,都要使用本技能——即使用户没说"升级"而只说"系统太老了""换个新系统但业务不能停"。In-place upgrade CentOS 7 to AlmaLinux 10 safely, single server or whole fleet.

FileKindSize
SKILL.mdskill md14.7 KB
assets/fleet-tracker.csvother617 B
assets/report-template.mdmarkdown850 B
references/app-compat.mdmarkdown14.5 KB
references/cloud-providers.mdmarkdown5.7 KB
references/fleet.mdmarkdown2.6 KB
references/kubernetes.mdmarkdown7.6 KB
references/pitfalls.mdmarkdown51.0 KB
references/recovery.mdmarkdown8.3 KB
scripts/backup.shexecutablescript5.0 KB
scripts/convert-el8.shexecutablescript7.6 KB
scripts/finalize.shexecutablescript12.5 KB
scripts/fleet_summary.pyexecutablescript7.9 KB
scripts/hop.shexecutablescript32.6 KB
scripts/k8s.shscript16.2 KB
scripts/known-hosts-fill.shscript2.7 KB
scripts/mysql-backup.shscript9.2 KB
scripts/nginx-static-build.shexecutablescript5.1 KB
scripts/post-audit.shexecutablescript12.4 KB
scripts/prep-el7.shexecutablescript13.5 KB
scripts/remote.pyexecutablescript9.7 KB
scripts/rescue-grubenv.shscript3.2 KB
scripts/state.shexecutablescript5.2 KB
scripts/survey.shexecutablescript28.6 KB
scripts/swap-openssh.shexecutablescript3.6 KB
scripts/verify.shexecutablescript5.3 KB

scripts/survey.sh

View on GitHub
#!/bin/bash
# survey.sh - 升级前只读勘察(EL7~EL10 通用,不修改任何东西)
# 用法: bash survey.sh > survey.txt     (末尾 RISK FLAGS 段是重点)
export LC_ALL=C
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
FLAGS=""
flag() { FLAGS="${FLAGS}[$1] $2"$'\n'; }   # 级别: BLOCKER / HIGH / MEDIUM / INFO
sec()  { echo; echo "===== $* ====="; }
has()  { command -v "$1" >/dev/null 2>&1; }
MAJOR=$(rpm -E %rhel 2>/dev/null); MAJOR=${MAJOR:-0}

sec OS
cat /etc/redhat-release 2>/dev/null; echo "kernel=$(uname -r) arch=$(uname -m) el=$MAJOR"
echo "hostname=$(hostname)"; uptime
VIRT=$(systemd-detect-virt 2>/dev/null)
VENDOR=$(cat /sys/class/dmi/id/sys_vendor 2>/dev/null); PRODUCT=$(cat /sys/class/dmi/id/product_name 2>/dev/null)
BIOSV=$(cat /sys/class/dmi/id/bios_vendor 2>/dev/null)
CLOUD=unknown
case "$VENDOR $PRODUCT $BIOSV" in *Alibaba*) CLOUD=aliyun;; *Amazon*|*EC2*) CLOUD=aws;; *OpenStack*|*HUAWEI*|*Huawei*) CLOUD=huawei;; esac
[ -d /usr/local/aegis ] || [ -d /usr/local/share/aliyun-assist ] && CLOUD=aliyun
[ -d /usr/local/uniagent ] || [ -d /CloudResetPwdUpdateAgent ] || [ -d /usr/local/hostguard ] && CLOUD=huawei
[ -e /etc/amazon ] || rpm -q amazon-ssm-agent >/dev/null 2>&1 && CLOUD=aws
echo "virt=$VIRT dmi='$VENDOR / $PRODUCT' cloud=$CLOUD"
[ -d /sys/firmware/efi ] && echo "boot=UEFI" || echo "boot=BIOS"
[ "$(uname -m)" != "x86_64" ] && flag BLOCKER "架构 $(uname -m):本技能只覆盖 x86_64"
[ "$MAJOR" = "7" ] && ! grep -q "7.9" /etc/redhat-release && flag INFO "当前非 7.9,ELevate 要求先 yum update 到 7.9"
OSID=$(. /etc/os-release 2>/dev/null; echo "$ID")
[ "$MAJOR" = 8 ] && [ "$OSID" != almalinux ] && flag BLOCKER "源系统是 $(cat /etc/redhat-release):ELevate 不支持以 CentOS 8 为源,先用 convert-el8.sh 转成 AlmaLinux 8,再 hop.sh 8→9→10"

sec CPU
grep -m1 "model name" /proc/cpuinfo; echo "nproc=$(nproc)"
MISS=""
for f in avx avx2 bmi1 bmi2 f16c fma abm movbe xsave; do grep -qw $f /proc/cpuinfo || MISS="$MISS $f"; done
if [ -n "$MISS" ]; then echo "x86-64-v3: NO (缺:$MISS)"; flag BLOCKER "CPU 不满足 x86-64-v3(缺$MISS):AlmaLinux 10 标准版无法运行,只能升到 9,或用 AlmaLinux 10 x86_64_v2 变体"
else echo "x86-64-v3: YES"; fi
# 华为云 c9 规格:AlmaLinux 10 内核在升级 initramfs 早期加载驱动时挂死(真实踩坑:2 台生产 + 2 台测试机;有无数据盘无关;
# 黑名单 ata_piix/ata_generic 实测无效;同一晚的 s3、早先的 c3 正常)
if grep -qiE "huawei" /sys/class/dmi/id/sys_vendor /sys/class/dmi/id/chassis_asset_tag 2>/dev/null; then
  HWSPEC=$(curl -s -m 3 http://169.254.169.254/openstack/latest/meta_data.json 2>/dev/null | grep -oE '"metering.resourcespeccode": *"[^"]*"' | sed -E 's/.*"([^"]*)"$/\1/')
  echo "huawei-flavor: ${HWSPEC:-unknown}"
  hwfam=${HWSPEC%%.*}
  if [ "$hwfam" = c9 ]; then
    flag HIGH "华为云 c9 规格($HWSPEC):AlmaLinux 10 内核在升级 initramfs 早期加载驱动时挂死(4 台实测,与数据盘无关,黑名单 ata_piix 无效),9→10 重启后卡住、VNC 抢不到 GRUB。只升到 AlmaLinux 9(pitfalls.md 9→10、recovery.md F)"
  elif grep -q "General Purpose Processor" /proc/cpuinfo; then
    flag INFO "华为云规格 ${HWSPEC:-unknown}(CPU 型号被隐藏,与 c9 同类):c9 上 EL10 内核会在升级 initramfs 里挂死,这个规格未验证,9→10 前先用同规格按需测试机试一次"
  fi
fi

sec MEMORY/DISK
free -h; df -hT -x tmpfs -x devtmpfs -x overlay 2>/dev/null
lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT 2>/dev/null
grep -vE "^\s*#|^\s*$" /etc/fstab
VARFREE=$(df -Pm /var | awk 'NR==2{print $4}'); BOOTFREE=$(df -Pm /boot | awk 'NR==2{print $4}')
[ "${VARFREE:-0}" -lt 5000 ] && flag HIGH "/var 可用 ${VARFREE}MB < 5GB:leapp 下载+容器环境需要 3~5GB,先清理(docker 无用镜像、日志、yum 缓存)"
[ "${BOOTFREE:-0}" -lt 250 ] && flag HIGH "/boot 可用 ${BOOTFREE}MB < 250MB:先删旧内核 (package-cleanup --oldkernels --count=1)"
grep -qE '^[^#]\S*\s+/(sys|proc|dev|run)/\S*\s+\S+\s+\S*bind' /etc/fstab && flag BLOCKER "fstab 里有 bind 到 /sys|/proc|/dev|/run 的挂载(如阿里云 set_ecs_special.sh 的 eth0-speed.mock):leapp 升级环境里挂不上直接进 emergency(shell 在串口,VNC 敲不了)。hop.sh up 会在 leapp upgrade 前自动注释,EL10 后恢复"
d=$(rpm -qa --qf '%{NAME}.%{ARCH}\n' 2>/dev/null | grep -vE '^(gpg-pubkey|kernel|kernel-devel|kernel-core|kernel-modules)\.' | sort | uniq -d | tr '\n' ' ')
[ -n "$d" ] && flag HIGH "rpm 库里有重复包(以前的 yum 事务没做完):$d→ yum update 会报 multilib/保护版本错误。先 package-cleanup --cleandupes;删不掉的旧副本 rpm -e --justdb --nodeps,再 update 并 reinstall 这些包"
# 注意:bash 4.2(EL7)里 $( ) 内不能写 case 的 "pattern)"(括号不配对会解析失败),所以放进函数
manual_listeners() {
  local p
  for p in $(ps -eo pid=); do
    grep -q "session-[0-9]*\.scope" /proc/$p/cgroup 2>/dev/null || continue
    grep -qE "docker|kubepods" /proc/$p/cgroup 2>/dev/null && continue
    case "$(cat /proc/$p/comm 2>/dev/null)" in bash|sh|sshd|sudo|su|sleep|tail|ps|grep|top|screen|tmux*) continue;; esac
    ss -lntp 2>/dev/null | grep -q "pid=$p," && cat /proc/$p/comm
  done
}
mp=$(manual_listeners | sort | uniq -c | tr '\n' ' ')
[ -n "$mp" ] && flag HIGH "有手工从 SSH 会话启动、正在监听端口的服务进程($mp):重启后不会自己起来。backup.sh 会记录命令行/目录/环境变量,升级后 finalize.sh manual-procs 生成 systemd 服务"
grep -qE "\s(nfs|cifs|glusterfs)\s" /etc/fstab && flag MEDIUM "fstab 有网络文件系统:升级环境里可能挂载失败,升级前临时注释 noauto/nofail"
lsblk -o FSTYPE 2>/dev/null | grep -qi vdo && flag HIGH "存在 VDO 设备:8→9 前需转换为 LVM-VDO"
lsblk -o FSTYPE 2>/dev/null | grep -qi btrfs && flag BLOCKER "使用 btrfs:EL8+ 不支持 btrfs"

sec PACKAGES
echo "total rpm: $(rpm -qa | wc -l)"
echo "--- 非发行版厂商的包 (vendor 非 CentOS/AlmaLinux/Red Hat):"
rpm -qa --qf '%{NAME}-%{VERSION}-%{RELEASE}.%{ARCH}\t%{VENDOR}\n' | grep -vE "CentOS|AlmaLinux|Red Hat|gpg-pubkey" | sort
echo "--- kernels:"; rpm -q kernel kernel-devel 2>/dev/null
KD=$(rpm -q kernel-devel 2>/dev/null | grep -vc "not installed"); [ "$KD" -gt 1 ] && flag HIGH "装了 $KD 个 kernel-devel(leapp 阻断项):只留当前内核的那个"
echo "--- repos:"; ls /etc/yum.repos.d/
for f in /etc/yum.repos.d/*.repo; do grep -hoE "^\[[^]]+\]" "$f"; done | sort | uniq -d | sed 's/^/DUP repo id: /'
EPV=$(grep -hE "^(baseurl|metalink|mirrorlist)=" /etc/yum.repos.d/epel.repo 2>/dev/null | grep -oE "epel/[0-9]+|epel-[0-9]+" | grep -oE "[0-9]+" | head -1)
[ -n "$EPV" ] && [ "$EPV" != "$MAJOR" ] && flag HIGH "epel.repo 指向 EPEL $EPV,但系统是 EL$MAJOR:转换/升级时会拉错包,先改成 EPEL $MAJOR"
rpm -qa | grep -qiE "^(php[0-9]*w|php7[0-9]w|php5[0-9]w)-" && flag HIGH "PHP 来自 webtatic(EL8+ 无此源):确认是否还用;要用就改 remi 或容器化"
rpm -qa | grep -qE "^mysql-community-libs" && flag HIGH "mysql-community-libs(-compat) 替换了 mariadb-libs:用 yum swap 换回 mariadb-libs,否则删它会连带删 postfix"
[ -e /etc/ld.so.conf.d/mysql-x86_64.conf ] && flag MEDIUM "自定义 ld.so.conf.d/mysql-x86_64.conf(leapp 会告警)"
rpm -qa | grep -qE "^(webtatic|ius|nginx-release|mysql[0-9]*-community-release|remi-release|elrepo-release)" && flag MEDIUM "存在第三方 *-release 源包,7→8 前卸载,EL10 后按需重新加"
REMI_ALL=$(rpm -qa --qf '%{NAME}|%{VENDOR}\n' 2>/dev/null | awk -F'|' 'tolower($2) ~ /remi/ {print $1}')
REMI_LIBS=$(echo "$REMI_ALL" | grep -vE '^(php|remi-release)' | tr '\n' ' ' | sed 's/ *$//')
[ -n "$REMI_LIBS" ] && [ "$MAJOR" = 7 ] && ! echo "$REMI_ALL" | grep -qE '^php' && flag MEDIUM "残留 Remi 兼容库(${REMI_LIBS})而宿主机已没有 Remi PHP:7→8 会在 dnf_package_download 和 EL8 的 libwebp/libzip 文件冲突(preupgrade 查不出来)。prep-el7.sh inhibitors 会删除"
rpm -qa | grep -qE "^(python2-|python-)" && [ "$MAJOR" = 8 ] && flag INFO "EL8 仍有 python2 包;EL9 起没有 python2"
HN=$(tr -d '[:space:]' < /etc/hostname 2>/dev/null); [ -z "$HN" ] && HN=$(hostname)
echo "$HN" | grep -q '[^A-Za-z0-9.-]' && flag MEDIUM "主机名 '$HN' 含下划线等非法字符:EL9 起 systemd 开机时会把这些字符去掉(实测 app_01 → app01),用 \$hostname 拼 ident 的监控 agent(categraf 等)会换成新目标、旧目标告警失联。升级后在 agent 配置里写死旧 ident,或改成合法主机名(见 pitfalls.md 8→9)"
# 不带版本号的 python shebang:EL8 起没有 /usr/bin/python,EL10 只有 python3 → 备份/巡检/日志轮转脚本静默失败
PYSH=""; for f in /usr/local/bin/* /usr/local/sbin/* /root/script*/* /root/bin/*; do
  [ -f "$f" ] || continue; head -1 "$f" 2>/dev/null | grep -qE '^#! *(/usr)?/bin/(env +)?python *$' && PYSH="$PYSH $f"
done
[ -n "$PYSH" ] && flag MEDIUM "脚本用不带版本的 python shebang(${PYSH# }):EL8 起没有 /usr/bin/python,cron/logrotate/systemd 调用会 203/EXEC 或 command not found(真实案例:GitLab 每日备份、磁盘巡检、nginx 日志轮转脚本)。确认兼容 python3 后把首行改成 #!/usr/bin/python3"
# 第三方 rpm 拥有的可执行文件被手工替换过:卸载/升级该包时会被删除或覆盖
MODF=$(rpm -qa --qf '%{NAME}|%{VENDOR}\n' 2>/dev/null | awk -F'|' 'tolower($2) !~ /(centos|red hat|almalinux|rocky)/ {print $1}' | while read -r p; do
  rpm -V --nomtime "$p" 2>/dev/null | awk '$1 ~ /5/ && $2 != "c" && $NF ~ /^\/(usr\/)?(local\/)?s?bin\//{print $NF}'; done | sort -u | tr '\n' ' ' | sed 's/ *$//')
[ -n "$MODF" ] && flag MEDIUM "第三方 rpm 拥有的可执行文件被手工替换过(${MODF}):卸载/升级该包时会被删除或覆盖(真实案例:/usr/bin/php 被换成调用 PHP 容器的包装脚本,卸载 remi php-cli 时一起被删)。先复制到 /usr/local/bin,卸载后再链接回来"
if rpm -q jenkins >/dev/null 2>&1; then
  JW=""; rpm -V jenkins 2>/dev/null | grep -q 'jenkins\.war$' && JW="(war 已被手工替换,rpm 版本不代表实际运行版本)"
  flag INFO "Jenkins rpm $(rpm -q --qf '%{VERSION}' jenkins)${JW}:老包是 SysV 启动 + 老签名(2.2xx 的 DSA/SHA1 签名是 8→9 阻断项)。见 app-compat.md 的 Jenkins:先换原生 systemd unit(环境变量与现在一致),8→9 前 rpm -e --justdb --noscripts jenkins"
fi
if command -v docker >/dev/null 2>&1 && docker info >/dev/null 2>&1; then
  GL=$(docker ps --format '{{.Names}} {{.Image}}' 2>/dev/null | awk 'tolower($2) ~ /gitlab\/gitlab-(ce|ee)|gitlab-(ce|ee):/ {print $1}' | while read -r n; do
    t=$(docker inspect -f '{{if .Config.StopTimeout}}{{.Config.StopTimeout}}{{end}}' "$n" 2>/dev/null); [ "${t:-0}" -lt 120 ] && echo "$n"; done | tr '\n' ' ')
  [ -n "$GL" ] && flag MEDIUM "GitLab omnibus 容器 ${GL}没设 stop_grace_period:dockerd 重启/关机只给 10s 就 SIGKILL,内置 PostgreSQL 留下 .s.PGSQL.5432.lock,之后一直 502(真实踩坑)。compose 里加 stop_grace_period: 5m;hop.sh reboot/post 会先优雅停止它"
fi
# 给 unit 设了内存上限但没设 OOMPolicy:EL9+ 的 systemd 默认 OOMPolicy=stop,内核杀掉一个进程就把整个 unit 停掉(连 mysqld_safe 这类守护一起),SysV 生成的 unit 又是 Restart=no → 服务一直挂着
MEMU=""
for f in /etc/systemd/system/*.service /etc/systemd/system/*.service.d/*.conf; do
  [ -e "$f" ] || continue
  grep -qE '^(MemoryLimit|MemoryMax)=' "$f" || continue
  u=${f#/etc/systemd/system/}; u=${u%%.d/*}
  grep -qsE '^OOMPolicy=' "/etc/systemd/system/$u" /etc/systemd/system/"$u".d/*.conf || MEMU="$MEMU $u"
done
[ -n "$MEMU" ] && flag MEDIUM "unit 设了 MemoryLimit/MemoryMax 但没设 OOMPolicy(${MEMU# }):EL7 上超限只杀一个进程,mysqld_safe 等守护会把它拉起来;EL9+ 默认 OOMPolicy=stop 会停掉整个 unit 且不重启(真实踩坑:MySQL 在每晚 mysqldump 时超 4G 被杀,一直停到第二天发版时才发现)。升级后改成 MemoryMax=(cgroup v2 下 MemoryLimit 已废弃)+ OOMPolicy=continue + Restart=on-failure,并按实际占用调大上限"

sec OPENSSH
rpm -q --qf '%{NAME}-%{VERSION}-%{RELEASE} vendor=%{VENDOR} sig=%{SIGPGP:pgpsig}\n' openssh-server 2>/dev/null
OSV=$(rpm -q --qf '%{VENDOR}' openssh-server 2>/dev/null)
case "$OSV" in *CentOS*|*AlmaLinux*|*"Red Hat"*) ;; *) flag BLOCKER "openssh-server 不是发行版官方包(vendor='$OSV',常见于自编译 9.x):先用 swap-openssh.sh 换回官方包,否则升级后 sshd 可能起不来=失联";; esac
[ -e /etc/rc.d/init.d/sshd ] && echo "sshd 使用 SysV 脚本 /etc/rc.d/init.d/sshd"
sshd -T 2>/dev/null | grep -E "^(port|permitrootlogin|passwordauthentication|pubkeyauthentication|usepam|clientaliveinterval|clientalivecountmax|ciphers|macs) "
grep -qE "^\s*PermitRootLogin" /etc/ssh/sshd_config || { [ "$(id -u)" = 0 ] && flag HIGH "sshd_config 未显式写 PermitRootLogin:EL9+ 默认 prohibit-password,root 密码登录会失联。显式写上当前需要的值"; }
CACM=$(sshd -T 2>/dev/null | awk '/^clientalivecountmax/{print $2}')
SSHV=$(ssh -V 2>&1 | grep -oE "OpenSSH_[0-9]+.[0-9]+" | cut -d_ -f2)
[ "$CACM" = "0" ] && awk -v v="$SSHV" 'BEGIN{exit !(v+0 < 8.2)}' && flag INFO "ClientAliveCountMax 0:OpenSSH<8.2 下空闲会话会被断,远程脚本要开 keepalive,长任务用 systemd-run"
grep -hs "ssh-dss" /root/.ssh/authorized_keys /home/*/.ssh/authorized_keys >/dev/null && flag BLOCKER "authorized_keys 含 ssh-dss (DSA) 公钥:EL9+ 不再支持,先换 ed25519/rsa 密钥"

sec SERVICES
echo "--- running:"; systemctl list-units --type=service --state=running --no-legend --no-pager | awk '{print $1}' | tr '\n' ' '; echo
echo "--- enabled:"; systemctl list-unit-files --state=enabled --no-legend --no-pager | awk '{print $1}' | tr '\n' ' '; echo
echo "--- failed:"; systemctl --failed --no-legend --no-pager
echo "--- SysV init scripts:"; ls /etc/rc.d/init.d/ 2>/dev/null | grep -vE "^(functions|README|netconsole|network)$"
echo "--- rc.local:"; RCL=$(grep -vE "^\s*#|^\s*$|touch /var/lock/subsys/local" /etc/rc.d/rc.local 2>/dev/null); echo "$RCL"
[ -n "$RCL" ] && flag MEDIUM "rc.local 里启动了东西($(echo "$RCL" | head -3 | tr '\n' ';')):升级后容易失效,改成 systemd unit"
echo "--- listening:"; ss -tulnp 2>/dev/null | awk 'NR>1{print $1, $5, $7}' | sort -u
systemctl is-active ntpd >/dev/null 2>&1 && flag INFO "ntpd 会被 leapp 迁移为 chronyd(自动)"

sec NETWORK
ip -br addr 2>/dev/null || ip addr; ip route; grep -v "^#" /etc/resolv.conf
echo "NetworkManager=$(systemctl is-active NetworkManager 2>/dev/null) network.service=$(systemctl is-active network 2>/dev/null)"
ls /etc/sysconfig/network-scripts/ifcfg-* 2>/dev/null
ls /etc/sysconfig/network-scripts/ifcfg-* >/dev/null 2>&1 && flag INFO "ifcfg 网络配置:9→10 前必须 nmcli connection migrate 转 keyfile(leapp 阻断项)"
[ "$(systemctl is-active NetworkManager 2>/dev/null)" != "active" ] && flag HIGH "NetworkManager 未运行(用的 network.service):EL8+ 以 NM 为准,升级前先切到 NM 并验证网络"
grep -rqs "^dhcp *= *dhclient" /etc/NetworkManager/ /usr/lib/NetworkManager/conf.d/ && flag HIGH "NetworkManager 配了 dhcp=dhclient(注意 /usr/lib 下也可能有不属于任何包的残留文件):EL10 没有 dhclient,是 9→10 阻断项;hop.sh answer 会处理"
for f in /etc/sysconfig/network-scripts/ifcfg-*; do [ -e "$f" ] || continue; case "$f" in *ifcfg-lo) continue;; esac; grep -q "^TYPE=" "$f" || flag MEDIUM "$(basename "$f") 没有 TYPE=:8→9 时 leapp 会报 unsupported device types(hop.sh answer 会用 nmcli connection migrate 处理)"; done
systemctl is-enabled iptables >/dev/null 2>&1 && flag INFO "iptables.service 已启用:EL9 起包名是 iptables-nft-services,8→9 会删除旧包;先看 /etc/sysconfig/iptables 里有没有真实规则"
echo "--- VPN / 隧道 / 访问路径:"
VPN=$(ps -eo comm,args 2>/dev/null | grep -iE "netbird|wireguard|wg-quick|openvpn|tailscale|zerotier|frpc|frps|strongswan|charon|softether|nps|npc" | grep -v grep)
ip -br link 2>/dev/null | grep -iE "wg|wt0|tun|tap|zt" ; echo "$VPN" | cut -c1-150
[ -n "$VPN" ] && flag BLOCKER "本机运行 VPN/隧道($(echo "$VPN" | awk '{print $1}' | sort -u | tr '\n' ' ')):先确认你是不是靠它连进来的!重启前绝不能 docker stop 它;准备好控制台 VNC/串口"

sec DOCKER/CONTAINERS
if has docker; then
  docker version --format 'server={{.Server.Version}} api={{.Server.APIVersion}}' 2>/dev/null
  docker info 2>/dev/null | grep -E "Storage Driver|Cgroup Driver|Cgroup Version|Docker Root Dir|Logging Driver"
  cat /etc/docker/daemon.json 2>/dev/null
  echo "--- containers (name | image | state | restart | network | compose dir):"
  for c in $(docker ps -aq 2>/dev/null); do
    docker inspect -f '{{.Name}} | {{.Config.Image}} | {{.State.Status}} | {{.HostConfig.RestartPolicy.Name}} | {{.HostConfig.NetworkMode}} | {{index .Config.Labels "com.docker.compose.project.working_dir"}}' "$c" 2>/dev/null
  done
  docker system df 2>/dev/null
  NOAUTO=$(for c in $(docker ps -q); do docker inspect -f '{{.Name}} {{.HostConfig.RestartPolicy.Name}}' $c; done | awk '$2=="no"||$2==""{print $1}' | tr '\n' ' ')
  [ -n "$NOAUTO" ] && flag HIGH "运行中但 restart=no 的容器(重启后不会自启):$NOAUTO"
  UNL=$(for c in $(docker ps -q); do docker inspect -f '{{.Name}} {{.HostConfig.RestartPolicy.Name}}' $c; done | awk '$2=="unless-stopped"{print $1}' | tr '\n' ' ')
  [ -n "$UNL" ] && flag HIGH "restart=unless-stopped 的容器:$UNL —— 重启前不要 docker stop 它们(手动停过就不会自启)"
  DV=$(docker version --format '{{.Server.Version}}' 2>/dev/null | cut -d. -f1); [ -n "$DV" ] && [ "$DV" -lt 20 ] 2>/dev/null && flag MEDIUM "Docker $DV.x 太老:leapp 会用 docker-ce 厂商源一起升级到新版(需要能访问 docker-ce 镜像源)"
  CV=$(docker-compose version --short 2>/dev/null); case "$CV" in 1.*) flag HIGH "docker-compose v1 ($CV):Docker≥25 下报 KeyError 'ContainerConfig' 并会把容器改名成 <id>_name,升级后换成 compose v2 包装脚本";; esac
  CG1=$(docker ps -a --format '{{.Image}}' | grep -iE "yearning|tidb|openjdk:8|java:8|jdk8|elasticsearch:[56]\.|mysql:5\.[0-6]" | sort -u | tr '\n' ' ')
  [ -n "$CG1" ] && flag MEDIUM "可能依赖 cgroup v1 的镜像:$CG1(EL9 默认、EL10 只有 cgroup v2;见 app-compat.md 的 cgroup v1 shim)"
else echo "docker: not installed"; fi
has podman && podman ps -a 2>/dev/null

sec KUBERNETES
for b in kubectl helm kubeadm kubelet crictl k9s kubeconfig kubeconf; do has $b && echo "$b: $(command -v $b) $($b version --client --short 2>/dev/null || $b version --short 2>/dev/null | head -1)"; done
ls -la /root/.kube/ 2>/dev/null; grep -rn "KUBECONFIG\|kubectl\|kubeconf" /root/.bashrc /root/.bash_profile /etc/profile.d/ 2>/dev/null
has kubectl && timeout 20 kubectl version 2>/dev/null | grep -i server
systemctl is-active kubelet >/dev/null 2>&1 && flag BLOCKER "本机是 Kubernetes 节点(kubelet 运行中):这是集群级变更,不能单台随手升。按 references/kubernetes.md 规划(访问路径是否在集群里、控制面入口、etcd、NFS 存储位置、能否整体中断),用 k8s.sh backup/pin/stop/fix/pods;换新节点也是选项"
systemctl is-active kubelet >/dev/null 2>&1 && flag INFO "K8s 节点到 EL10 后只有 cgroup v2:pod 里 8u372 / 11.0.16 之前的老 JDK 认不出容器内存限制,会按宿主机内存算堆而 OOMKilled → 恢复后用 k8s.sh jvm-maxram"
[ "$MAJOR" -lt 10 ] && rpm -q kubectl >/dev/null 2>&1 && flag INFO "kubectl 是 rpm 安装($(rpm -q kubectl)):升级后换成与集群版本匹配的 pkgs.k8s.io 源"

sec RUNTIMES/DATABASES
has java && java -version 2>&1 | head -1; rpm -qa 'java-*-openjdk*' 2>/dev/null | head
rpm -qa 'java-*-openjdk*' | grep -qE "java-(1\.8\.0|11|17)-openjdk" && flag MEDIUM "宿主机 Java 8/11/17:EL10 只有 21/25,宿主机直接跑的 Java 程序要改用容器或自带 JDK"
has python2 && echo "python2: $(python2 -V 2>&1)"; has python3 && echo "python3: $(python3 -V 2>&1)"
grep -rls "python2\|/usr/bin/python " /etc/cron* /var/spool/cron 2>/dev/null | sed 's/^/cron uses python2: /'
PIPN=0
for d in /usr/lib/python2.7/site-packages /usr/lib64/python2.7/site-packages /usr/lib/python3.6/site-packages /usr/lib64/python3.6/site-packages /usr/lib/python3.9/site-packages /usr/lib64/python3.9/site-packages; do
  [ -d "$d" ] || continue
  for f in "$d"/*; do [ -e "$f" ] || continue; rpm -qf "$f" >/dev/null 2>&1 || { echo "UNOWNED $f"; PIPN=$((PIPN+1)); }; done
done
[ "$PIPN" -gt 0 ] && flag BLOCKER "有 $PIPN 个 pip 装的、不属于任何 rpm 的 site-packages 文件:升级时和 rpm 文件冲突(cpio rename)会让升级在 initramfs 里半途失败!EL7 上运行 prep-el7.sh pipfix(EL8/9 上 hop.sh prep 会自动挪走)"
has php && php -v 2>/dev/null | head -1
for p in mysqld mariadbd postgres postmaster mongod redis-server elasticsearch; do pgrep -x $p >/dev/null 2>&1 && echo "DB process: $p ($(readlink /proc/$(pgrep -xo $p)/exe 2>/dev/null))"; done
# cron 里调用 mysqldump 的备份脚本:升级后 /usr/bin/mysqldump 变成 8.4(不认 MYSQL_PWD),而老脚本常 >/dev/null 2>&1 静默失败
MDS=$(for u in $(cut -d: -f1 /etc/passwd); do crontab -l -u "$u" 2>/dev/null; done | grep -v '^\s*#' | grep -oE '/[^ ;|>]+\.sh' | sort -u | while read -r s; do [ -f "$s" ] && grep -lq mysqldump "$s" 2>/dev/null && echo "$s"; done | tr '\n' ' ')
[ -n "$MDS" ] && flag INFO "cron 里有 mysqldump 备份脚本(${MDS}):升级前后都确认最近一次备份文件结尾有 '-- Dump completed'、各库都有文件(真实案例:全库备份每晚在一个坏库中止,持续几个月没人发现)。EL10 的 mysqldump 是 8.4,不认 MYSQL_PWD。可换成 scripts/mysql-backup.sh(逐库校验 + 监控指标)"
# 发布/备份脚本里 tar 的 --exclude 写在目录参数之后:EL9+ 的 GNU tar ≥1.34 报错退出(2)且 exclude 不生效
TARRE='(^|[;&|[:space:]"])tar[[:space:]]+-?[a-zA-Z]*c[a-zA-Z]*f[[:space:]]+[^[:space:]-][^[:space:]]*[[:space:]]+[^[:space:]-][^[:space:]]*[^#]*--exclude'
TARX=$( { for u in $(cut -d: -f1 /etc/passwd); do crontab -l -u "$u" 2>/dev/null; done | grep -v '^\s*#' | grep -oE '/[^ ;|>]+\.sh'
  for d in /root/script /root/scripts /data/script /data/scripts /opt/script /opt/scripts; do [ -d "$d" ] && find "$d" -maxdepth 3 -type f -name '*.sh' 2>/dev/null; done; } | sort -u | head -500 | while read -r s; do
  [ -f "$s" ] && grep -vE '^[[:space:]]*#' "$s" 2>/dev/null | grep -qE "$TARRE" && echo "$s"; done | head -10 | tr '\n' ' ')
[ -n "$TARX" ] && flag MEDIUM "脚本里 tar 的 --exclude 写在目录参数后面(${TARX}):EL9+ 的 tar 1.34 会报 'options were used after any non-optional arguments' 并以 2 退出,exclude 也不生效(set -e/Jenkins 发布直接失败,或备份把 runtime/.git 全打包)。把选项挪到目录前面;跑发布任务的 CI(Jenkins 等)里的同类命令也要查"
# Tomcat:<Context docBase="xxx.war"> 指向 appBase 里的 war,同时 autoDeploy/deployOnStartup 开着 → 同一个 war 被部署两次
TCD=""
for sx in /data/*/conf/server.xml /opt/*/conf/server.xml /usr/local/*/conf/server.xml /usr/share/tomcat*/conf/server.xml; do
  [ -e "$sx" ] || continue
  grep -vE '^\s*<!--' "$sx" | grep -qE '<Context[^>]*docBase="[^"/][^"]*\.war"' || continue
  grep -qE 'deployOnStartup="false"' "$sx" && continue
  TCD="$TCD $sx"
done
[ -n "$TCD" ] && flag MEDIUM "Tomcat 的 <Context docBase=\"*.war\"> 指向 appBase 内的 war 且自动部署开着(${TCD# }):同一个 war 会以 / 和 /<名字> 部署两次,Jenkins 这类会起两个实例共用同一个 HOME,重启后谁先起来不确定(真实案例:升级后重启 ROOT 起不来,报 key cannot be empty or null,/ 返回 404)。Host 上加 autoDeploy=\"false\" deployOnStartup=\"false\";不要挪走那个 war(它就是 ROOT)"
# CI/发布机:known_hosts 里只记录 ECDSA/RSA 的主机,到 EL10 后会 Host key verification failed(crypto-policies 固定了 HostKeyAlgorithms)
if [ "$(rpm -E %rhel 2>/dev/null)" -lt 10 ] 2>/dev/null && [ -s /root/.ssh/known_hosts ]; then
  KHN=$(awk '$1 !~ /^\|/ && $1 !~ /^#/ {t[$1]=t[$1]" "$2} END{for(h in t) if (t[h] !~ /ssh-ed25519/) n++; print n+0}' /root/.ssh/known_hosts)
  [ "$KHN" -gt 0 ] && flag INFO "/root/.ssh/known_hosts 里 $KHN 个主机没有 ed25519 记录:本机若用 ssh/scp/rsync 发布(Jenkins 等),升级到 EL10 后这些主机会 Host key verification failed。到 10 后运行 known-hosts-fill.sh 补齐"
fi
rpm -qa | grep -qE "^(mariadb-server|postgresql-server|mysql-server)-" && flag BLOCKER "宿主机装了 rpm 版数据库服务($(rpm -qa | grep -E '^(mariadb-server|postgresql-server|mysql-server)-' | tr '\n' ' ')):每跳都会大版本升级,数据目录格式变化。先逻辑备份(mysqldump/pg_dumpall),制定每跳 mysql_upgrade/pg_upgrade 方案"

sec CUSTOM BINARIES(运行中、非 rpm 安装的可执行文件及其 EL7 库依赖)
for pid in $(ls /proc | grep -E '^[0-9]+$'); do readlink /proc/$pid/exe 2>/dev/null; done | grep -v " (deleted)" | sort -u | while read -r exe; do
  [ -f "$exe" ] || continue
  rpm -qf "$exe" >/dev/null 2>&1 && continue
  case "$exe" in /var/lib/docker/*|/var/lib/containerd/*|/proc/*) continue;; esac
  libs=$(ldd "$exe" 2>/dev/null | grep -oE "(libssl\.so\.10|libcrypto\.so\.10|libssl\.so\.1\.0[^ ]*|libpcre\.so\.1|libpython2[^ ]*|libmysqlclient\.so\.18|libboost[^ ]*\.so\.1\.53[^ ]*|libicu[^ ]*\.so\.50|libffi\.so\.6|libreadline\.so\.6|libnsl\.so\.1|not found)" | sort -u | tr '\n' ' ')
  echo "$exe  ${libs:+[EL7-libs: $libs]}"
  [ "$MAJOR" -le 8 ] && ldd "$exe" 2>/dev/null | grep -q 'libcrypt\.so\.1 ' && echo "$exe" >> /tmp/.survey_libcrypt1
  [ -n "$libs" ] && echo "$exe|$libs" >> /tmp/.survey_custom_bins
done
if [ -s /tmp/.survey_custom_bins ]; then
  while IFS='|' read -r e l; do flag HIGH "自编译程序 $e 依赖 EL7 库($l):升级后会起不来。方案:静态编译(nginx-static-build.sh)或改用新系统官方包"; done < /tmp/.survey_custom_bins
  rm -f /tmp/.survey_custom_bins
fi
if [ -s /tmp/.survey_libcrypt1 ]; then
  flag INFO "非 rpm 程序依赖 libcrypt.so.1($(sort -u /tmp/.survey_libcrypt1 | tr '\n' ' ')):EL9 起在 libxcrypt-compat 里,hop.sh prep 会让 leapp 一起装上"
  rm -f /tmp/.survey_libcrypt1
fi
for n in $(pgrep -a nginx 2>/dev/null | grep master | grep -oE "/[^ ]*nginx( |$)" | sort -u); do
  [ -x "$n" ] || { echo "--- $n: 在容器里运行(宿主机无此文件),见 DOCKER 段"; continue; }
  echo "--- $n -V:"; $n -V 2>&1
  $n -T 2>/dev/null | grep -nE "^\s*ssl\s+on;" | head -3 | grep -q . && flag MEDIUM "nginx 配置用了 'ssl on;':nginx ≥1.25.1 已删除,迁移到新版时改成 'listen 443 ssl;'"
  $n -T 2>/dev/null | grep -qE "^\s*add_header\s+['\"]?[A-Za-z0-9-]+:['\"]?\s" && flag INFO "nginx 有 add_header 的头名带冒号(如 'Access-Control-Allow-Credentials:'):HTTP/2 客户端(EL8+ 的 curl、部分浏览器)会 PROTOCOL_ERROR;升级前就存在,告诉业务方"
  echo "--- server_name:"; $n -T 2>/dev/null | grep -E "^\s*server_name" | sort -u
done
pgrep -a httpd 2>/dev/null | head -2

sec CRON/USERS
for u in $(cut -d: -f1 /etc/passwd); do crontab -l -u "$u" 2>/dev/null | grep -vE "^\s*#|^\s*$" | sed "s/^/[$u] /"; done
ls /etc/cron.d/; awk -F: '$3>=1000 && $3<65000 {print "user:", $1, $7}' /etc/passwd

sec CLOUD AGENTS
for d in /usr/local/aegis /usr/local/share/aliyun-assist /usr/local/cloudmonitor /usr/local/uniagent /usr/local/hostguard /opt/oss/servicemgr/ICAgent /CloudResetPwdUpdateAgent /CloudResetPwdAgent/opt/aws /snap/amazon-ssm-agent /usr/bin/amazon-ssm-agent /opt/aws/amazon-cloudwatch-agent; do [ -e "$d" ] && echo "agent: $d"; done
echo "cloud-init: $(rpm -q cloud-init 2>/dev/null) status=$(cloud-init status 2>/dev/null)"
ls /usr/lib/python2.7/site-packages 2>/dev/null | grep -q "cloud_init.*egg" && flag MEDIUM "cloud-init 是 pip/egg 安装的 py2 版本:升级后会坏,安装新系统的 cloud-init 包或禁用(AWS 慎重,见 cloud-providers.md)"

sec MAIL/MISC
has postconf && postconf -n 2>/dev/null | grep -E "hash:|btree:" && flag INFO "postfix 用了 hash:/btree: 表:EL10 无 libdb,升级后改 lmdb: 并 newaliases/postmap"
getenforce 2>/dev/null
lsmod | grep -E "^(floppy|pata_acpi|btrfs|nft_compat|ip_set) " | awk '{print "loaded module:", $1}'
lsmod | grep -qE "^(floppy|pata_acpi) " && flag HIGH "加载了 floppy/pata_acpi(7→8 阻断项):rmmod 并加入 /etc/modprobe.d 黑名单"

echo
echo "################ RISK FLAGS ################"
if [ -z "$FLAGS" ]; then echo "(none)"; else
  for lv in BLOCKER HIGH MEDIUM INFO; do printf "%s" "$FLAGS" | grep "^\[$lv\]"; done
fi
echo "############################################"
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall