故障排查
修复：Fix PHP Fpm Kill Long Backup
Multisite 备份「PHP Fpm Kill Long Backup」的原因与修复。
自由的
Fix PHP Fpm Kill Long Backup 的修复
When you hit php fpm kill long backup, start with the Multisite Migrate job log: last completed chunk, PHP/memory messages, HTTP 502/504, disk/inode errors, or cloud auth failures. Multisite archives are larger and table-heavier than single-site jobs, so shared hosts interrupt long PHP requests — Jobs in Chunks exist specifically for that.
Check tmp/disk free space, upload_max_filesize for imports, S3/Drive/Dropbox/SFTP credentials for Plus uploads, and security plugins that block admin-ajax. After restore issues, verify siteurl/home, rewrite rules, domain mapping (sunrise.php), object-cache drop-ins, and CDN caches. Reproduce once on Staging before changing production limits.
「PHP Fpm Kill Long Backup」先看日志诊断
Separate backup failures from post-restore URL/media/login problems. Capture job percent, last log line, HTTP status, hosting class, and whether a local archive already exists. That evidence decides whether you raise PHP limits, free disk, fix cloud auth, or only re-upload.
For “php fpm kill long backup”, avoid restarting from scratch if a chunked job can resume. Lost locks, cleared tmp, or a new job ID are common reasons a progress bar jumps back to 0%.
贴合 Multisite 现实的修复
Raise PHP time/memory enough for safety, but keep chunked jobs — shared hosts can still kill long workers. Free disk and inodes before large network archives. Allow admin-ajax through WAF or security plugins. On Plus, re-test Drive/Dropbox/S3-compatible/SFTP credentials and quotas.
Free handles core backup/restore; Plus covers BYO cloud; Pro covers remote import, empty-server installer, Staging, and split/extract. Pick the tool that matches the failure — do not upgrade tiers hoping a cloud auth typo will disappear.
归档完成后
If the local archive is good, you can restore even when cloud upload failed. Fix destination auth/quota, then re-upload, or restore from local. On Staging, confirm HTTPS, mapping, media, and logins before production.
If only one subsite looks wrong after restore, check that site’s tables and sites/{id}/ uploads path before re-running the whole network job.
保留、加密与支持卫生
For fix php fpm kill long backup, keep enough archive generations to roll back a bad deploy or failed migration. Optional .venc AES-256 encryption on Multisite Migrate archives protects copies at rest when you store offsite on Plus destinations. Document credentials, schedule windows, and who may run Network Admin jobs.
Separate schedule failures from restore failures in your runbook so on-call staff do not rebuild a finished archive unnecessarily.
为何重要
日志优先诊断
判断失败来自 PHP、磁盘、数据库、云还是 URL 重写。
默认修复：Jobs in Chunks
续传，而不是把大型 Multisite 任务从零重跑。
恢复侧检查
把备份失败与恢复后的 URL/媒体/登录问题分开。
便于支持的证据
知道该为支持保留哪些日志行与环境事实。
步骤
-
1记录失败点
在 Network Admin 或子站点执行此步骤，并在 Staging 上验证结果。
-
2应用对应修复
在 Network Admin 或子站点执行此步骤，并在 Staging 上验证结果。
-
3以 chunks 重跑
在 Network Admin 或子站点执行此步骤，并在 Staging 上验证结果。
-
4Confirm local archive integrity（本地化）
在 Network Admin 或子站点执行此步骤，并在 Staging 上验证结果。
-
5校验恢复路径
在 Network Admin 或子站点执行此步骤，并在 Staging 上验证结果。
常见问题
关于此 Multisite 备份主题的简要解答。仍有问题？打开文档或联系支持。
问题：Should I raise PHP limits forever?
Multisite Migrate 答复：Free/Plus/Pro、分块任务、自带云与 Staging 恢复演练。
问题：Why did the job restart at 0%?
通常是锁丢失、tmp 被清或新的任务 ID。重跑前检查锁与磁盘。
问题：Cloud upload failed but local archive exists?
修好目标认证/配额后再上传，或用本地归档恢复。
问题：Is “php fpm kill long backup” usually backup or restore?
Multisite Migrate 答复：Free/Plus/Pro、分块任务、自带云与 Staging 恢复演练。