Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

tikv: refine region-cache log and metric (#10589) #10667

Merged
merged 2 commits into from
Jun 3, 2019
Merged

tikv: refine region-cache log and metric (#10589) #10667

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1e47ea7
tikv: refine region-cache log and metric (#10589)
lysu May 27, 2019
9fa272a
Merge branch 'release-3.0' into cp-region-cache-metric-30
zz-jason Jun 3, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 25 additions & 11 deletions store/tikv/region_cache.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,13 +38,14 @@ const (
)

var (
tikvRegionCacheCounterWithDropRegionFromCacheOK = metrics.TiKVRegionCacheCounter.WithLabelValues("drop_region_from_cache", "ok")
tikvRegionCacheCounterWithGetRegionByIDOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region_by_id", "ok")
tikvRegionCacheCounterWithGetRegionByIDError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region_by_id", "err")
tikvRegionCacheCounterWithGetRegionOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region", "ok")
tikvRegionCacheCounterWithGetRegionError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region", "err")
tikvRegionCacheCounterWithGetStoreOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_store", "ok")
tikvRegionCacheCounterWithGetStoreError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_store", "err")
tikvRegionCacheCounterWithInvalidateRegionFromCacheOK = metrics.TiKVRegionCacheCounter.WithLabelValues("invalidate_region_from_cache", "ok")
tikvRegionCacheCounterWithSendFail = metrics.TiKVRegionCacheCounter.WithLabelValues("send_fail", "ok")
tikvRegionCacheCounterWithGetRegionByIDOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region_by_id", "ok")
tikvRegionCacheCounterWithGetRegionByIDError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region_by_id", "err")
tikvRegionCacheCounterWithGetRegionOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region", "ok")
tikvRegionCacheCounterWithGetRegionError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_region", "err")
tikvRegionCacheCounterWithGetStoreOK = metrics.TiKVRegionCacheCounter.WithLabelValues("get_store", "ok")
tikvRegionCacheCounterWithGetStoreError = metrics.TiKVRegionCacheCounter.WithLabelValues("get_store", "err")
)

const (
Expand Down Expand Up @@ -121,6 +122,7 @@ func (r *Region) checkRegionCacheTTL(ts int64) bool {

// invalidate invalidates a region, next time it will got null result.
func (r *Region) invalidate() {
tikvRegionCacheCounterWithInvalidateRegionFromCacheOK.Inc()
atomic.StoreInt64(&r.lastAccess, invalidatedLastAccessTime)
}

Expand Down Expand Up @@ -359,13 +361,18 @@ func (c *RegionCache) findRegionByKey(bo *Backoffer, key []byte, isEndKey bool)
}

// OnSendFail handles send request fail logic.
func (c *RegionCache) OnSendFail(bo *Backoffer, ctx *RPCContext, scheduleReload bool) {
func (c *RegionCache) OnSendFail(bo *Backoffer, ctx *RPCContext, scheduleReload bool, err error) {
tikvRegionCacheCounterWithSendFail.Inc()
r := c.getCachedRegionWithRLock(ctx.Region)
if r != nil {
c.switchNextPeer(r, ctx.PeerIdx)
if scheduleReload {
r.scheduleReload()
}
logutil.Logger(bo.ctx).Info("switch region peer to next due to send request fail",
zap.Stringer("current", ctx),
zap.Bool("needReload", scheduleReload),
zap.Error(err))
}
}

Expand Down Expand Up @@ -457,7 +464,6 @@ func (c *RegionCache) InvalidateCachedRegion(id RegionVerID) {
if cachedRegion == nil {
return
}
tikvRegionCacheCounterWithDropRegionFromCacheOK.Inc()
cachedRegion.invalidate()
}

Expand All @@ -473,14 +479,23 @@ func (c *RegionCache) UpdateLeader(regionID RegionVerID, leaderStoreID uint64, c

if leaderStoreID == 0 {
c.switchNextPeer(r, currentPeerIdx)
logutil.Logger(context.Background()).Info("switch region peer to next due to NotLeader with NULL leader",
zap.Int("currIdx", currentPeerIdx),
zap.Uint64("regionID", regionID.GetID()))
return
}

if !c.switchToPeer(r, leaderStoreID) {
logutil.Logger(context.Background()).Debug("regionCache: cannot find peer when updating leader",
logutil.Logger(context.Background()).Info("invalidate region cache due to cannot find peer when updating leader",
zap.Uint64("regionID", regionID.GetID()),
zap.Int("currIdx", currentPeerIdx),
zap.Uint64("leaderStoreID", leaderStoreID))
r.invalidate()
} else {
logutil.Logger(context.Background()).Info("switch region leader to specific leader due to kv return NotLeader",
zap.Uint64("regionID", regionID.GetID()),
zap.Int("currIdx", currentPeerIdx),
zap.Uint64("leaderStoreID", leaderStoreID))
}
}

Expand Down Expand Up @@ -712,7 +727,6 @@ func (c *RegionCache) OnRegionEpochNotMatch(bo *Backoffer, ctx *RPCContext, curr
if needInvalidateOld {
cachedRegion, ok := c.mu.regions[ctx.Region]
if ok {
tikvRegionCacheCounterWithDropRegionFromCacheOK.Inc()
cachedRegion.invalidate()
}
}
Expand Down
8 changes: 4 additions & 4 deletions store/tikv/region_cache_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -239,7 +239,7 @@ func (s *testRegionCacheSuite) TestSendFailedButLeaderNotChange(c *C) {
c.Assert(len(ctx.Meta.Peers), Equals, 3)

// send fail leader switch to 2
s.cache.OnSendFail(s.bo, ctx, false)
s.cache.OnSendFail(s.bo, ctx, false, nil)
ctx, err = s.cache.GetRPCContext(s.bo, loc.Region)
c.Assert(err, IsNil)
c.Assert(ctx.Peer.Id, Equals, s.peer2)
Expand Down Expand Up @@ -267,7 +267,7 @@ func (s *testRegionCacheSuite) TestSendFailedInHibernateRegion(c *C) {
c.Assert(len(ctx.Meta.Peers), Equals, 3)

// send fail leader switch to 2
s.cache.OnSendFail(s.bo, ctx, false)
s.cache.OnSendFail(s.bo, ctx, false, nil)
ctx, err = s.cache.GetRPCContext(s.bo, loc.Region)
c.Assert(err, IsNil)
c.Assert(ctx.Peer.Id, Equals, s.peer2)
Expand Down Expand Up @@ -303,13 +303,13 @@ func (s *testRegionCacheSuite) TestSendFailedInMultipleNode(c *C) {
c.Assert(len(ctx.Meta.Peers), Equals, 3)

// send fail leader switch to 2
s.cache.OnSendFail(s.bo, ctx, false)
s.cache.OnSendFail(s.bo, ctx, false, nil)
ctx, err = s.cache.GetRPCContext(s.bo, loc.Region)
c.Assert(err, IsNil)
c.Assert(ctx.Peer.Id, Equals, s.peer2)

// send 2 fail leader switch to 3
s.cache.OnSendFail(s.bo, ctx, false)
s.cache.OnSendFail(s.bo, ctx, false, nil)
ctx, err = s.cache.GetRPCContext(s.bo, loc.Region)
c.Assert(err, IsNil)
c.Assert(ctx.Peer.Id, Equals, peer3)
Expand Down
2 changes: 1 addition & 1 deletion store/tikv/region_request.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -172,7 +172,7 @@ func (s *RegionRequestSender) onSendFail(bo *Backoffer, ctx *RPCContext, err err
}
}

s.regionCache.OnSendFail(bo, ctx, s.needReloadRegion(ctx))
s.regionCache.OnSendFail(bo, ctx, s.needReloadRegion(ctx), err)

// Retry on send request failure when it's not canceled.
// When a store is not available, the leader of related region should be elected quickly.
Expand Down