forked from mtusi2026/mtusi2026lab4
93 lines
4.4 KiB
INI
93 lines
4.4 KiB
INI
# Client -> HAProxy -> api x N. Replicas are discovered through Docker DNS and checked actively.
|
|
#
|
|
# Timing contract (keep in sync with compose.yaml and app settings):
|
|
# check period inter 1s, timeout check 1s
|
|
# marked DOWN after fall 2 failed checks -> hung/draining replica leaves the pool in <= 3 s
|
|
# marked UP after rise 2 good checks -> new replica gets traffic ~2 s after it is ready
|
|
# DNS refresh every 1 s (hold valid 1s); a removed container disappears within ~1-2 s
|
|
# crashed process connect refused -> request redispatched to another replica at once
|
|
# app DRAIN_SECONDS=4 > 3 s, so a replica receiving SIGTERM is out of rotation before it closes its socket.
|
|
|
|
global
|
|
log stdout format raw local0 info
|
|
maxconn 2048
|
|
# Runtime API for evidence (scripts/lb-status.sh); loopback inside the container only.
|
|
stats socket ipv4@127.0.0.1:9999 level operator
|
|
|
|
resolvers docker
|
|
nameserver dns 127.0.0.11:53
|
|
accepted_payload_size 8192
|
|
resolve_retries 3
|
|
timeout resolve 1s
|
|
timeout retry 1s
|
|
hold valid 1s
|
|
hold other 1s
|
|
hold refused 1s
|
|
hold nx 1s
|
|
hold timeout 1s
|
|
hold obsolete 0s
|
|
|
|
defaults
|
|
mode http
|
|
log global
|
|
option httplog
|
|
option dontlognull
|
|
option forwardfor
|
|
timeout connect 1s
|
|
timeout client 30s
|
|
timeout server 10s
|
|
# Body arrival for a 10 MiB attachment; JSON posts finish well under this.
|
|
timeout http-request 30s
|
|
timeout http-keep-alive 5s
|
|
timeout queue 2s
|
|
timeout check 1s
|
|
# A request whose connection to a replica failed is sent to another replica.
|
|
retries 2
|
|
option redispatch 1
|
|
|
|
frontend public
|
|
bind :8080
|
|
monitor-uri /lb/health
|
|
# Attachment uploads are raw bytes, up to 10 MiB. A larger Content-Length is
|
|
# rejected here, before the body is read or an API replica is contacted.
|
|
# The body matches the API error envelope; request_id echoes a safe client
|
|
# X-Request-Id and otherwise stays "lb", like the other gateway errors.
|
|
acl upload_too_large req.hdr_val(content-length) gt 10485760
|
|
acl safe_request_id req.hdr(X-Request-Id) -m reg ^[A-Za-z0-9._:-]{1,128}$
|
|
http-request set-var(txn.reqid) req.hdr(X-Request-Id) if upload_too_large safe_request_id
|
|
http-request set-var(txn.reqid) str(lb) if upload_too_large !safe_request_id
|
|
http-request return status 413 content-type application/json lf-string "{\"error\":{\"code\":\"payload_too_large\",\"message\":\"Файл больше 10 МиБ\",\"request_id\":\"%[var(txn.reqid)]\"}}" hdr X-Request-Id %[var(txn.reqid)] hdr X-Instance-Id %[str(lb)] if upload_too_large
|
|
|
|
# Message POST carries a mandatory Idempotency-Key, so replaying it on another replica is safe.
|
|
# Other POSTs (users, chats, sessions) are retried only when the connection was never established.
|
|
acl message_post method POST
|
|
acl message_post_path path_reg ^/api/v1/chats/[^/]+/messages$
|
|
http-request disable-l7-retry if message_post !message_post_path
|
|
|
|
default_backend api
|
|
|
|
backend api
|
|
# Round robin, no stickiness: any replica serves any session from PostgreSQL.
|
|
# Attachment PUT to S3 can outlast the default 10 s server timeout under concurrent load.
|
|
acl attachment_upload method POST
|
|
acl attachment_upload_path path_reg ^/api/v1/chats/[^/]+/attachments$
|
|
http-request set-timeout server 30s if attachment_upload attachment_upload_path
|
|
balance roundrobin
|
|
option httpchk
|
|
http-check send meth GET uri /health/ready ver HTTP/1.1 hdr Host lb
|
|
http-check expect status 200
|
|
retry-on conn-failure empty-response response-timeout
|
|
http-error status 503 content-type application/json string "{\"error\":{\"code\":\"unavailable\",\"message\":\"No API replica is available\",\"request_id\":\"lb\"}}"
|
|
http-error status 502 content-type application/json string "{\"error\":{\"code\":\"internal_error\",\"message\":\"Bad gateway\",\"request_id\":\"lb\"}}"
|
|
http-error status 504 content-type application/json string "{\"error\":{\"code\":\"internal_error\",\"message\":\"Gateway timeout\",\"request_id\":\"lb\"}}"
|
|
default-server check inter 1s fall 2 rise 2 resolvers docker resolve-prefer ipv4 init-addr none
|
|
server-template api 1-10 api:8080
|
|
|
|
# Read-only stats for evidence; not published, reachable only inside the container.
|
|
frontend stats
|
|
bind 127.0.0.1:8404
|
|
stats enable
|
|
stats uri /stats
|
|
stats refresh 5s
|
|
http-request set-log-level silent
|