remotes: add mirrorlist for round-robin + failover across mirrors (rpm/deb/apk) (#121)
## Why
OS package remotes (rpm/deb/apk) fetch many small files and benefit from spreading upstream load across mirrors and surviving a mirror outage. A remote may now set a **`mirrorlist`** of additional upstream base URLs. The effective upstream pool is **`[base_url] + mirrorlist`**, which the shared proxy engine load-balances **round-robin** and, on a network error/timeout/5xx, **fails over** to the next mirror before returning an error. Selection happens in the engine, so it works for every provider that reaches upstream.
**Backward compatible:** `base_url` stays a plain string (providers read it unchanged), and a remote with **no mirrorlist behaves exactly as today** (single attempt, same error path).
## How
- `models.Remote.Mirrorlist` (`[]string`, `json:"mirrorlist,omitempty"`) + `UpstreamPool()` = `[base_url] + mirrorlist`.
- `ValidateMirrorlist`: a non-empty mirrorlist is allowed **only** when `repo_type==remote` **and** `package_type ∈ {rpm, deb, alpine}`; each entry must be an http/https URL. Enforced in the v2 create/update handlers (400 otherwise); `base_url` stays required for remotes.
- Persist the mirrorlist in a new additive `mirrorlist TEXT[]` column (`remoteCols`/`scanRemote`/`CreateRemote`/`UpdateRemote`); the `base_url` column is unchanged.
- Engine keeps a per-remote round-robin cursor over the pool; the fetch/head/revalidate upstream calls run in a failover loop that narrows the remote to one selected mirror per attempt. Only network errors and 5xx fail over (404/403/… return as-is). The circuit breaker stays keyed per remote and trips only after all mirrors fail.
## Scope
Round-robin + failover only, restricted to **remote rpm/deb/apk** repos. Least-connections and a per-remote strategy selector are a **follow-up PR**.
## Tests
- Unit: model JSON round-trip + validation gating (rejected on non-rpm/deb/apk and on local, accepted on rpm/deb/apk, bad URL rejected), engine round-robin/failover/no-mirrorlist-unchanged, DB mirrorlist round-trip. `make test` (`go test -race`) green.
- Docker acceptance (`e2e-docker`, `dockere2e` tag, wired into `docker-e2e.sh`): round-robin distribution across two mock upstreams, failover past a dead primary, no-mirrorlist regression, and a **real `dnf` makecache + install** through a two-mirror rpm remote whose `base_url` is dead. All four pass locally.
Reviewed-on: #121
Co-authored-by: unkin-agent <unkin-agent@unkin.net>
Co-committed-by: unkin-agent <unkin-agent@unkin.net>
This commit was merged in pull request #121.
This commit is contained in:
@@ -11,6 +11,8 @@ import (
|
||||
"log/slog"
|
||||
"net/http"
|
||||
"strings"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
"time"
|
||||
|
||||
"git.unkin.net/unkin/artifactapi/internal/cache"
|
||||
@@ -35,6 +37,10 @@ type Engine struct {
|
||||
cas *storage.CAS
|
||||
circuit *CircuitBreaker
|
||||
accessLog chan database.AccessLogEntry
|
||||
// rrCounters holds a per-remote round-robin cursor (remoteName ->
|
||||
// *atomic.Uint64) used to rotate the starting mirror across upstream base
|
||||
// URLs. Distribution is per-replica and approximate, which is fine.
|
||||
rrCounters sync.Map
|
||||
}
|
||||
|
||||
func NewEngine(db *database.DB, c *cache.Redis, s *storage.S3) *Engine {
|
||||
@@ -222,7 +228,30 @@ func (e *Engine) Head(ctx context.Context, remote models.Remote, path string, pr
|
||||
return e.headUpstream(ctx, remote, path, prov)
|
||||
}
|
||||
|
||||
// headUpstream issues an upstream HEAD, load-balancing across the remote's base
|
||||
// URLs and failing over to the next mirror on a network error or 5xx.
|
||||
func (e *Engine) headUpstream(ctx context.Context, remote models.Remote, path string, prov provider.Provider) (*HeadResult, error) {
|
||||
order := e.baseURLAttemptOrder(remote)
|
||||
if len(order) == 0 {
|
||||
return nil, &ProxyError{Status: http.StatusBadGateway, Message: "no upstream base_url configured"}
|
||||
}
|
||||
var lastErr error
|
||||
for i, url := range order {
|
||||
result, err := e.headUpstreamOnce(ctx, withBaseURL(remote, url), path, prov)
|
||||
if err == nil {
|
||||
return result, nil
|
||||
}
|
||||
lastErr = err
|
||||
if i < len(order)-1 && shouldFailover(err) {
|
||||
slog.Warn("upstream HEAD failed, failing over", "remote", remote.Name, "base_url", url, "error", err)
|
||||
continue
|
||||
}
|
||||
return nil, err
|
||||
}
|
||||
return nil, lastErr
|
||||
}
|
||||
|
||||
func (e *Engine) headUpstreamOnce(ctx context.Context, remote models.Remote, path string, prov provider.Provider) (*HeadResult, error) {
|
||||
url := prov.UpstreamURL(remote, path)
|
||||
|
||||
authHeaders, err := prov.AuthHeaders(ctx, remote)
|
||||
@@ -277,7 +306,31 @@ func (e *Engine) headUpstream(ctx context.Context, remote models.Remote, path st
|
||||
return &HeadResult{ContentType: contentType, Size: resp.ContentLength, Source: "remote"}, nil
|
||||
}
|
||||
|
||||
// fetchFromUpstream fetches an artifact from upstream, load-balancing across the
|
||||
// remote's base URLs and failing over to the next mirror on a network error or
|
||||
// 5xx before returning an error.
|
||||
func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, path string, prov provider.Provider, class Classification, ttl time.Duration, clientHeaders http.Header) (*FetchResult, error) {
|
||||
order := e.baseURLAttemptOrder(remote)
|
||||
if len(order) == 0 {
|
||||
return nil, &ProxyError{Status: http.StatusBadGateway, Message: "no upstream base_url configured"}
|
||||
}
|
||||
var lastErr error
|
||||
for i, url := range order {
|
||||
result, err := e.fetchFromUpstreamOnce(ctx, withBaseURL(remote, url), path, prov, class, ttl, clientHeaders)
|
||||
if err == nil {
|
||||
return result, nil
|
||||
}
|
||||
lastErr = err
|
||||
if i < len(order)-1 && shouldFailover(err) {
|
||||
slog.Warn("upstream fetch failed, failing over", "remote", remote.Name, "base_url", url, "error", err)
|
||||
continue
|
||||
}
|
||||
return nil, err
|
||||
}
|
||||
return nil, lastErr
|
||||
}
|
||||
|
||||
func (e *Engine) fetchFromUpstreamOnce(ctx context.Context, remote models.Remote, path string, prov provider.Provider, class Classification, ttl time.Duration, clientHeaders http.Header) (*FetchResult, error) {
|
||||
url := prov.UpstreamURL(remote, path)
|
||||
|
||||
authHeaders, err := prov.AuthHeaders(ctx, remote)
|
||||
@@ -454,7 +507,31 @@ func (e *Engine) serveFromStore(ctx context.Context, remote models.Remote, path
|
||||
}, nil
|
||||
}
|
||||
|
||||
// checkUpstream issues a conditional upstream HEAD (If-None-Match), load
|
||||
// balancing across the remote's base URLs and failing over to the next mirror on
|
||||
// a network error or 5xx.
|
||||
func (e *Engine) checkUpstream(ctx context.Context, remote models.Remote, path, etag string, prov provider.Provider) (bool, error) {
|
||||
order := e.baseURLAttemptOrder(remote)
|
||||
if len(order) == 0 {
|
||||
return false, &ProxyError{Status: http.StatusBadGateway, Message: "no upstream base_url configured"}
|
||||
}
|
||||
var lastErr error
|
||||
for i, url := range order {
|
||||
notModified, err := e.checkUpstreamOnce(ctx, withBaseURL(remote, url), path, etag, prov)
|
||||
if err == nil {
|
||||
return notModified, nil
|
||||
}
|
||||
lastErr = err
|
||||
if i < len(order)-1 && shouldFailover(err) {
|
||||
slog.Warn("upstream revalidation failed, failing over", "remote", remote.Name, "base_url", url, "error", err)
|
||||
continue
|
||||
}
|
||||
return false, err
|
||||
}
|
||||
return false, lastErr
|
||||
}
|
||||
|
||||
func (e *Engine) checkUpstreamOnce(ctx context.Context, remote models.Remote, path, etag string, prov provider.Provider) (bool, error) {
|
||||
url := prov.UpstreamURL(remote, path)
|
||||
|
||||
req, err := http.NewRequestWithContext(ctx, http.MethodHead, url, nil)
|
||||
@@ -649,3 +726,45 @@ func isNetworkError(err error) bool {
|
||||
var ue *UpstreamError
|
||||
return errors.As(err, &ue)
|
||||
}
|
||||
|
||||
// baseURLAttemptOrder returns the ordered upstream base URLs to try for a single
|
||||
// request, drawn from the remote's pool ([base_url] + mirrorlist). A multi-mirror
|
||||
// remote starts at the next round-robin position and advances linearly for
|
||||
// failover; a remote with no mirrorlist yields exactly [base_url], preserving the
|
||||
// original single-attempt behavior.
|
||||
func (e *Engine) baseURLAttemptOrder(remote models.Remote) []string {
|
||||
urls := remote.UpstreamPool()
|
||||
if len(urls) <= 1 {
|
||||
return urls
|
||||
}
|
||||
v, _ := e.rrCounters.LoadOrStore(remote.Name, new(atomic.Uint64))
|
||||
start := int(v.(*atomic.Uint64).Add(1) - 1)
|
||||
ordered := make([]string, len(urls))
|
||||
for i := range urls {
|
||||
ordered[i] = urls[(start+i)%len(urls)]
|
||||
}
|
||||
return ordered
|
||||
}
|
||||
|
||||
// withBaseURL narrows a remote's active BaseURL to a single selected mirror so
|
||||
// providers (UpstreamURL/AuthHeaders/RewriteResponse) operate on exactly that
|
||||
// upstream for this attempt.
|
||||
func withBaseURL(remote models.Remote, url string) models.Remote {
|
||||
remote.BaseURL = url
|
||||
remote.Mirrorlist = nil
|
||||
return remote
|
||||
}
|
||||
|
||||
// shouldFailover reports whether an upstream attempt error is worth retrying
|
||||
// against the next mirror: network errors/timeouts and upstream 5xx responses.
|
||||
// Definitive statuses (404/403/401/...) are returned to the caller unchanged.
|
||||
func shouldFailover(err error) bool {
|
||||
if isNetworkError(err) {
|
||||
return true
|
||||
}
|
||||
var pe *ProxyError
|
||||
if errors.As(err, &pe) {
|
||||
return pe.Status >= 500
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user