"fail2ban protects you from brute force"
It logs and bans — after the attempt hits sshd. With a 3-retry / 10-min ban, a botnet rotating 5000 IPs still gets 15k guesses/hour through. fail2ban barely dents that.
Real fix: kill password auth entirely.
—
—
— move port off 22 to cut log noise
Now the bans are cosmetic and you sleep. fail2ban is a seatbelt, not a lock.
Try it tonight.
It logs and bans — after the attempt hits sshd. With a 3-retry / 10-min ban, a botnet rotating 5000 IPs still gets 15k guesses/hour through. fail2ban barely dents that.
Real fix: kill password auth entirely.
—
PasswordAuthentication no—
PubkeyAuthentication yes— move port off 22 to cut log noise
Now the bans are cosmetic and you sleep. fail2ban is a seatbelt, not a lock.
Try it tonight.
Killed the :00 load spike by staggering cron
Just did this on a $6 box running 15 sites. Every wp-cron and backup fired at the top of the hour, load avg hit 9.0 for ~90 seconds, sites timing out.
Moved real crons off wp-cron and spread them:
— disabled wp-cron, added per-site system cron at odd minutes (
— backups jittered with
Peak load avg fell from 9.0 to 2.1. No more clustered timeouts. The jitter one-liner alone flattened the backup pileup.
Try it tonight.
Just did this on a $6 box running 15 sites. Every wp-cron and backup fired at the top of the hour, load avg hit 9.0 for ~90 seconds, sites timing out.
Moved real crons off wp-cron and spread them:
— disabled wp-cron, added per-site system cron at odd minutes (
7,23,41 * * * *)— backups jittered with
sleep $((RANDOM \% 300)) at the top of the jobPeak load avg fell from 9.0 to 2.1. No more clustered timeouts. The jitter one-liner alone flattened the backup pileup.
Try it tonight.
Forwarded from AFF.TOP - про арбитраж трафика и CPA рынок!
This media is not supported in your browser
VIEW IN TELEGRAM
Meta ограничивает расходы на токены для сотрудников
Meta ввела внутренние лимиты на использование ИИ из-за резкого роста расходов: в 2026 году только на сотрудников заложены миллиарды долларов, а общий бюджет на ИИ-инфраструктуру оценивается в 130–145 млрд. Вывод простой: даже у Big Tech ИИ перестал быть бесплатной игрушкой и требует жёсткого контроля затрат.
➡️ Читайте на сайте: https://aff.top/blog/meta-ogranichivaet-raskhody-na-tokeny-dlia-sotrudnikov
🧠 Ещё больше инсайтов → в канале AFF.top
Meta ввела внутренние лимиты на использование ИИ из-за резкого роста расходов: в 2026 году только на сотрудников заложены миллиарды долларов, а общий бюджет на ИИ-инфраструктуру оценивается в 130–145 млрд. Вывод простой: даже у Big Tech ИИ перестал быть бесплатной игрушкой и требует жёсткого контроля затрат.
➡️ Читайте на сайте: https://aff.top/blog/meta-ogranichivaet-raskhody-na-tokeny-dlia-sotrudnikov
🧠 Ещё больше инсайтов → в канале AFF.top
Forwarded from AFF.TOP - про арбитраж трафика и CPA рынок!
This media is not supported in your browser
VIEW IN TELEGRAM
Claude Cowork, Claude Design объединили в один Claude
➡️ Читайте на сайте: https://aff.top/blog/claude-cowork-claude-design-obedinili-v-odin-claude
🧠 Ещё больше инсайтов → в канале AFF.top
➡️ Читайте на сайте: https://aff.top/blog/claude-cowork-claude-design-obedinili-v-odin-claude
🧠 Ещё больше инсайтов → в канале AFF.top
Forwarded from AffPapa! Клуб спящих бизнесменов! Потрачено!
This media is not supported in your browser
VIEW IN TELEGRAM
🔥 Приватные консультации по запускам Google ads и FB.
Масштабное обновление материала на сентябрь,без воды и паблика,свежий пак информации для опытных баеров(техничка,разбан,модерация,
связки,масштабирование и т.д)
Полный пак:
https://t.me/googleadsroi/164558
Отзывы:
https://t.me/+jnxGdX6GbjgxZTQx
Аккаунты гугл адс:
https://t.me/+VCIrjC36UiYyYjM0
Мой контакт:@TRAFF3
гарант+По промокоду( #affpapa ) скидка -10% на все услуги.
Please open Telegram to view this post
VIEW IN TELEGRAM
"More PHP-FPM workers = handle more traffic"
Nope. Each
Math it:
Fewer workers that never swap beat many that do. A queued request is fine; a swapping box is dead.
Try it tonight.
Nope. Each
pm.max_children eats 30-60MB. Set it to 50 on a 2GB box and you'll swap-thrash into a death spiral the second traffic spikes.Math it:
(RAM - 512MB for MySQL/nginx) / avg_worker_MB. On 2GB that's ~20 children, not 50.Fewer workers that never swap beat many that do. A queued request is fine; a swapping box is dead.
Try it tonight.
nginx vs Caddy on a $5 box
Just moved 3 sites off Caddy back to nginx. Caddy's auto-TLS is gorgeous for a single site you forget about. But the second you need fine-grained
— Caddy: 1 site, set-and-forget, you hate config files
— nginx: many sites, PHP-FPM, you want cache control
Caddy ate 80MB idle. nginx sits at 12MB.
Try it tonight.
Just moved 3 sites off Caddy back to nginx. Caddy's auto-TLS is gorgeous for a single site you forget about. But the second you need fine-grained
fastcgi_cache, micro-cache zones, or rate-limit buckets, nginx wins.— Caddy: 1 site, set-and-forget, you hate config files
— nginx: many sites, PHP-FPM, you want cache control
Caddy ate 80MB idle. nginx sits at 12MB.
Try it tonight.
Forwarded from AFF.TOP - про арбитраж трафика и CPA рынок!
This media is not supported in your browser
VIEW IN TELEGRAM
Microsoft планирует вставлять рекламу в игры
➡️ Читайте на сайте: https://aff.top/blog/microsoft-planiruet-vstavliat-reklamu-v-igry
🧠 Ещё больше инсайтов → в канале AFF.top
➡️ Читайте на сайте: https://aff.top/blog/microsoft-planiruet-vstavliat-reklamu-v-igry
🧠 Ещё больше инсайтов → в канале AFF.top
Forwarded from AffPapa! Клуб спящих бизнесменов! Потрачено!
This media is not supported in your browser
VIEW IN TELEGRAM
Совсем скоро запуск ШЕСТОГО проекта на RU GEO от создателей APEX, EVA, KUSH, BANDA и LEEBET!
Please open Telegram to view this post
VIEW IN TELEGRAM
"Set nginx worker_processes to your core count"
Half-right and outdated. Just use
Default 768 means 1 vCPU caps at ~768 concurrent. Bump it:
—
—
Also raise the OS limit or nginx silently caps. One $5 vCPU served 8k concurrent for me after this.
Try it tonight.
Half-right and outdated. Just use
worker_processes auto; — it reads cores itself. The real bottleneck nobody mentions is worker_connections.Default 768 means 1 vCPU caps at ~768 concurrent. Bump it:
—
worker_connections 4096;—
worker_rlimit_nofile 8192;Also raise the OS limit or nginx silently caps. One $5 vCPU served 8k concurrent for me after this.
Try it tonight.
rsync vs restic for backups
rsync mirrors files fast and you can read the result with any tool — but it's not real versioning; an oops overwrites your only copy.
restic does deduplicated, encrypted, snapshot backups straight to cheap S3/B2.
— Quick local mirror to a second disk: rsync
— Versioned offsite backups, ransomware-proof: restic
B2 storage runs ~$6/TB/month, restic dedupes so you pay for almost nothing. Mirror locally, snapshot offsite. Try it tonight.
rsync mirrors files fast and you can read the result with any tool — but it's not real versioning; an oops overwrites your only copy.
restic does deduplicated, encrypted, snapshot backups straight to cheap S3/B2.
— Quick local mirror to a second disk: rsync
— Versioned offsite backups, ransomware-proof: restic
B2 storage runs ~$6/TB/month, restic dedupes so you pay for almost nothing. Mirror locally, snapshot offsite. Try it tonight.