open-falcon-archive / graph Goto Github PK

我们上报了一个数值，python 写的，value 大概是 79.999909123972 这个样子。基本不变。
但是看图的时候，graph 的 y 座标全部写的 80.000，曲线看起来波动很大（实际上极差不超过 0.0000001）。

应该可以限制绘图的最小精度？

想问下，我如果想修改历史数据存储粒度为10s钟的点存储7天，应该怎么修改rrdtool.go文件

该如何定位慢的原因？有最佳实践吗？
一次性查看30台机器的数据，counter视图。大约3-4秒，有时要多次刷新才出现。endpoint视图快一些。
如何提高查询速度。

环境是四台Graph组成的集群，已经稳定运行一段时间，go是1.6.2，系统是centos 6，ssd硬盘

但是今天发现日志中出现大量的 errno: 0x023a, str:opening error 问题，大概每台有千行左右，四台graph都出现这样的问题，而且影响了出图，重启后又正常。

而且这个应该与 #17 和 #12 有些不同

没有像 #17 中那样伴随着 errno: 0panic: runtime error: cgo argument has Go pointer to Go pointer

也没有像 #12 中那样一直报问题，版本也比较新，除了 #25 其他commit都有，而且部署的是实体机，没用docker

对于此问题也是没有一点头绪，请问能否解答分析一下问题出现的原因，如何避免呢？

@laiwei @niean @hitripod

graph内存使用越来越大，不断增长，最后导致oomk，如何解决，谢谢~

"rrd": {
    "storage": "/home/work/data/6070"
}

这段storage的配置，是否可以配置多个？
如果可以，该怎么配置？
不可以，若需要多个存储目录，该怎么配置？启动多个graph实例吗？还是咋整

在我们线上部署graph后，rrd数据存储量非常大，结合rrd存储的相关概念，由于不懂开发语言，线上已经部署，所以有以下疑问紧急请教以下：
1.按照文档的解释，我得理解是：graph将采集的数据通过rrd的方式存储，同一个metric存一个rrd文件，rrd文件大小事固定的，相同metric在不同时间点的数据采集上来都在同一个rrd文件中，直至rrd环被写满，默认情况是五年，但是我们线上的graph数据存储目录下的rrd文件数量在不断的增长，只是磁盘使用情况非常的巨大。
2.线上监控的项目中包括nginx和mysql，插件用得open-falcon官方推荐的，这些采集的metric数量比较大，会不会产生他们的metric疯狂创建rrd文件的情况
非常感谢！！！

MySQL只需要配一台吗？graph集群里的多个graph连接的是同一个mysql？
类似的问题，多个judge连接的也是同一个redis实例？

程序运行一会出这个错误

errno: 0x023a, str:opening error
errno: 0panic: runtime error: cgo argument has Go pointer to Go pointer

goroutine 23 [running]:
panic(0x90e0a0, 0xc8203d24e0)
/usr/local/go/src/runtime/panic.go:464 +0x3e6
github.com/open-falcon/rrdlite.(_Updater).update(0xc820032de8, 0xc82039b3c0, 0x6, 0x8, 0x0, 0x0)
/home/work/workspace/src/github.com/open-falcon/rrdlite/rrd_c.go:70 +0x100
github.com/open-falcon/rrdlite.(_Updater).Update(0xc820032de8, 0x0, 0x0, 0x0, 0x0, 0x0)
/home/work/workspace/src/github.com/open-falcon/rrdlite/rrd.go:130 +0x27f
github.com/open-falcon/graph/rrdtool.update(0xc8203ae9b0, 0x46, 0xc8203a9020, 0x6, 0x6, 0x0, 0x0)
/home/work/workspace/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:117 +0x45e
github.com/open-falcon/graph/rrdtool.flushrrd(0xc8203ae9b0, 0x46, 0xc8203a9020, 0x6, 0x6, 0x0, 0x0)
/home/work/workspace/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:142 +0x1a2
github.com/open-falcon/graph/rrdtool.ioWorker()
/home/work/workspace/src/github.com/open-falcon/graph/rrdtool/sync_disk.go:95 +0x38f
created by github.com/open-falcon/graph/rrdtool.Start
/home/work/workspace/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:54 +0x24f

我在push data时，设定的timestamp比如是
tp=int(time.mktime(time.strptime('2017-07-08 00:10:00', '%Y-%m-%d %H:%M:%S')))
tp是1499443800，给定的value=93
我在调用/graph/history时候，设定"start": 1499150029, "end": 1499505624,返回的结果里面却没有这个时间戳，
[
{
"endpoint": "test-endpoint",
"counter": "test-metric/stats=cw",
"dstype": "GAUGE",
"step": 86400,
"Values": [
{
"timestamp": 1499212800,
"value": null
},
{
"timestamp": 1499299200,
"value": null
},
{
"timestamp": 1499385600,
"value": 93
},
{
"timestamp": 1499472000,
"value": null
}
]
}
]
值为93的时间戳显示的是1499385600，为什么会不一样呢？

没有错误日志，如何定位原因

现在group模块对metric在同一个timestamp的数据只保留收到的第一份，但业务上我们需要实现类似update的功能，数据有了变化会多次发送timestamp相同的数据。简单把下面代码中的continue去掉会到导致程序错误，有什么好的建议吗？

    proc.GraphRpcRecvCnt.Incr()

    // To Graph
    first := store.GraphItems.First(key)
    if first != nil && items[i].Timestamp <= first.Timestamp {
        continue
    }
    store.GraphItems.PushFront(key, items[i], checksum, cfg)

    // To Index
    index.ReceiveItem(items[i], checksum)

    // To History
    store.AddItem(checksum, items[i])

curl -vsq localhost:6071/history/

• Curl_http_done: called premature == 0
• Empty reply from server
• Connection #0 to host localhost left intact

$ cd ~/falcon-graph/
$ ./control tail

2017/05/12 18:08:51 server.go:2753: http: panic serving [::1]:59985: runtime error: index out of range
goroutine 20788 [running]:
net/http.(*conn).serve.func1(0xc4200a2c80)
	/usr/local/go/src/net/http/server.go:1721 +0xd0
panic(0x4474900, 0x4720b30)
	/usr/local/go/src/runtime/panic.go:489 +0x2cf
github.com/open-falcon/graph/http.configProcRoutes.func3(0x46fb0a0, 0xc4202582a0, 0xc42000aa00)
	/Users/youzhengchuan/go/src/github.com/open-falcon/graph/http/proc_http.go:30 +0x35c
net/http.HandlerFunc.ServeHTTP(0x450fe28, 0x46fb0a0, 0xc4202582a0, 0xc42000aa00)
	/usr/local/go/src/net/http/server.go:1942 +0x44
net/http.(*ServeMux).ServeHTTP(0x472f380, 0x46fb0a0, 0xc4202582a0, 0xc42000aa00)
	/usr/local/go/src/net/http/server.go:2238 +0x130
net/http.serverHandler.ServeHTTP(0xc4201ee000, 0x46fb0a0, 0xc4202582a0, 0xc42000aa00)
	/usr/local/go/src/net/http/server.go:2568 +0x92
net/http.(*conn).serve(0xc4200a2c80, 0x46fba20, 0xc420224680)
	/usr/local/go/src/net/http/server.go:1825 +0x612
created by net/http.(*Server).Serve
	/usr/local/go/src/net/http/server.go:2668 +0x2ce

有时候报这个错误 errno: 0x023d, str:mmap error 不知道什么原因，没有详细说明。
已经更新到最新的代码，golang 1.7.4 （docker官方仓库golang:lastest版本） build的。

最近做tomcat的监控数据push到agent时，发现DashBoard上图标全是空白的，然后看falcon的日志，发现graph的app.log总是在报错：
[mysql] 2016/05/09 14:51:04 packets.go:33: unexpected EOF
[mysql] 2016/05/09 14:51:04 packets.go:124: write tcp 127.0.0.1:3306: broken pipe
[mysql] 2016/05/09 14:51:04 packets.go:124: write tcp 127.0.0.1:3306: broken pipe
[mysql] 2016/05/09 14:51:18 packets.go:33: unexpected EOF
[mysql] 2016/05/09 14:51:18 packets.go:124: write tcp 127.0.0.1:3306: broken pipe

是不是这里写数据库失败，导致DashBoard上都是空白 啊

[root@yafei graph]# ./control start
start ok, pid=32122
[root@yafei graph]# ./control tail
2016/06/15 10:04:20 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/06/15 10:04:20 db.go:22: g.InitDB, get db conn fail Error 1045: Access denied for user 'root'@'localhost' (using password: NO)
2016/06/15 10:06:09 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/06/15 10:06:09 db.go:22: g.InitDB, get db conn fail Error 1193: Unknown system variable 'parsseTime'
2016/06/15 10:29:11 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/06/15 10:29:11 db.go:22: g.InitDB, get db conn fail Error 1193: Unknown system variable 'parsseTime'
2016/06/15 10:36:27 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/06/15 10:36:27 db.go:22: g.InitDB, get db conn fail Error 1193: Unknown system variable 'parsseTime'
2016/06/15 10:36:46 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/06/15 10:36:46 db.go:22: g.InitDB, get db conn fail Error 1193: Unknown system variable 'parsseTime'

例如配置成:
/a/b/c

如果没有b目录，那么不会生成rrd文件，且不会报错。

store/storage.go 文件中

func (this *GraphItemMap) First(key string) *cmodel.GraphItem {
    this.RLock()
    defer this.RUnlock()
    idx := int(hashKey(key)) % this.Size
    L, ok := this.A[idx][key]

• hashKey()函数返回值为uint32的，如果返回的值为3061298526，计算出来的idx为负数，graph直接
  panic。
  神人，帮着看一下^_^

root@fdc055cabfe3:/home/work/open-falcon# cat logs/graph.log
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error
errno: 0x023a, str:opening error

使用docker镜像默认配置部署的，dashboard 全部是空白没数据，发现这个错误信息, 不明白什么意思。

"rrd": {
"storage": "/home/work/open-falcon/data/6070"
},
这个目录下是空的，感觉是写数据打开文件失败？，但root启动有写入权限的

用的是最新版本的代码（gitversion 2d27dc2）
长期运行一直很稳定，今天凌晨意外崩溃了。。
可能和当时磁盘IO偏高有关
日志显示这样：

fatal error: concurrent map read and map write

goroutine 61213088 [running]:
runtime.throw(0x86725a, 0x21)
/usr/local/go/src/runtime/panic.go:566 +0x95 fp=0xc438085250 sp=0xc438085230
runtime.mapaccess2(0x7e6e60, 0xc42000d650, 0xc4399b9ac0, 0xc4380852e0, 0xc4380852e0)
/usr/local/go/src/runtime/hashmap.go:340 +0x249 fp=0xc438085298 sp=0xc438085250
reflect.mapaccess(0x7e6e60, 0xc42000d650, 0xc4399b9ac0, 0xc42000d650)
/usr/local/go/src/runtime/hashmap.go:1008 +0x3f fp=0xc4380852d0 sp=0xc438085298
reflect.Value.MapIndex(0x7e6e60, 0xc420088220, 0x195, 0x7c6f20, 0xc4399b9ac0, 0x98, 0x7dcc40, 0xc4399b9af0, 0x94)
/usr/local/go/src/reflect/value.go:1040 +0x128 fp=0xc438085358 sp=0xc4380852d0
encoding/json.(*mapEncoder).encode(0xc42012c208, 0xc42032e0b0, 0x7e6e60, 0xc420088220, 0x195, 0x100)
/usr/local/go/src/encoding/json/encode.go:646 +0x4f0 fp=0xc4380854b0 sp=0xc438085358
encoding/json.(*mapEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x7e6e60, 0xc420088220, 0x195, 0x100)
/usr/local/go/src/encoding/json/encode.go:662 +0x64 fp=0xc4380854f0 sp=0xc4380854b0
encoding/json.(*structEncoder).encode(0xc420418ba0, 0xc42032e0b0, 0x83a420, 0xc4200881c0, 0x199, 0x100)
/usr/local/go/src/encoding/json/encode.go:601 +0x253 fp=0xc438085650 sp=0xc4380854f0
encoding/json.(*structEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x83a420, 0xc4200881c0, 0x199, 0xc420080100)
/usr/local/go/src/encoding/json/encode.go:615 +0x64 fp=0xc438085690 sp=0xc438085650
encoding/json.(*ptrEncoder).encode(0xc42012c210, 0xc42032e0b0, 0x827500, 0xc4200881c0, 0x16, 0x820100)
/usr/local/go/src/encoding/json/encode.go:742 +0xe3 fp=0xc4380856d0 sp=0xc438085690
encoding/json.(*ptrEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x827500, 0xc4200881c0, 0x16, 0xc420080100)
/usr/local/go/src/encoding/json/encode.go:747 +0x64 fp=0xc438085710 sp=0xc4380856d0
encoding/json.(*encodeState).reflectValue(0xc42032e0b0, 0x827500, 0xc4200881c0, 0x16, 0xc420080100)
/usr/local/go/src/encoding/json/encode.go:307 +0x82 fp=0xc438085748 sp=0xc438085710
encoding/json.interfaceEncoder(0xc42032e0b0, 0x7dcc40, 0xc42e1b02a0, 0x194, 0x7d0100)
/usr/local/go/src/encoding/json/encode.go:573 +0xdb fp=0xc438085788 sp=0xc438085748
encoding/json.(*arrayEncoder).encode(0xc42012c1a8, 0xc42032e0b0, 0x7bfa40, 0xc4416efaa0, 0x97, 0x100)
/usr/local/go/src/encoding/json/encode.go:723 +0xee fp=0xc4380857e0 sp=0xc438085788
encoding/json.(*arrayEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x7bfa40, 0xc4416efaa0, 0x97, 0x100)
/usr/local/go/src/encoding/json/encode.go:730 +0x64 fp=0xc438085820 sp=0xc4380857e0
encoding/json.(*sliceEncoder).encode(0xc42012c1b0, 0xc42032e0b0, 0x7bfa40, 0xc4416efaa0, 0x97, 0x7b0100)
/usr/local/go/src/encoding/json/encode.go:697 +0xc1 fp=0xc438085860 sp=0xc438085820
encoding/json.(*sliceEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x7bfa40, 0xc4416efaa0, 0x97, 0xc4416e0100)
/usr/local/go/src/encoding/json/encode.go:709 +0x64 fp=0xc4380858a0 sp=0xc438085860
encoding/json.(*encodeState).reflectValue(0xc42032e0b0, 0x7bfa40, 0xc4416efaa0, 0x97, 0xc4416e0100)
/usr/local/go/src/encoding/json/encode.go:307 +0x82 fp=0xc4380858d8 sp=0xc4380858a0
encoding/json.interfaceEncoder(0xc42032e0b0, 0x7dcc40, 0xc4416efcb0, 0x94, 0x100)
/usr/local/go/src/encoding/json/encode.go:573 +0xdb fp=0xc438085918 sp=0xc4380858d8
encoding/json.(*structEncoder).encode(0xc420418a20, 0xc42032e0b0, 0x802880, 0xc4416efca0, 0x99, 0x800100)
/usr/local/go/src/encoding/json/encode.go:601 +0x253 fp=0xc438085a78 sp=0xc438085918
encoding/json.(*structEncoder).(encoding/json.encode)-fm(0xc42032e0b0, 0x802880, 0xc4416efca0, 0x99, 0xc4416e0100)
/usr/local/go/src/encoding/json/encode.go:615 +0x64 fp=0xc438085ab8 sp=0xc438085a78
encoding/json.(*encodeState).reflectValue(0xc42032e0b0, 0x802880, 0xc4416efca0, 0x99, 0x100)
/usr/local/go/src/encoding/json/encode.go:307 +0x82 fp=0xc438085af0 sp=0xc438085ab8
encoding/json.(*encodeState).marshal(0xc42032e0b0, 0x802880, 0xc4416efca0, 0x100, 0x0, 0x0)
/usr/local/go/src/encoding/json/encode.go:280 +0xb8 fp=0xc438085b28 sp=0xc438085af0
encoding/json.Marshal(0x802880, 0xc4416efca0, 0x20, 0xc4204d2401, 0xc4416efca0, 0xc42bde3bd0, 0x41131b)
/usr/local/go/src/encoding/json/encode.go:145 +0x8f fp=0xc438085b70 sp=0xc438085b28
github.com/open-falcon/graph/http.RenderJson(0xc3d3e0, 0xc4287aa410, 0x802880, 0xc4416efca0)
/usr/local/gopath/src/github.com/open-falcon/graph/http/http.go:32 +0x39 fp=0xc438085bd0 sp=0xc438085b70
github.com/open-falcon/graph/http.RenderDataJson(0xc3d3e0, 0xc4287aa410, 0x7bfa40, 0xc4416efaa0)
/usr/local/gopath/src/github.com/open-falcon/graph/http/http.go:42 +0xb4 fp=0xc438085c28 sp=0xc438085bd0
github.com/open-falcon/graph/http.configProcRoutes.func2(0xc3d3e0, 0xc4287aa410, 0xc4336d50e0)
/usr/local/gopath/src/github.com/open-falcon/graph/http/proc_http.go:20 +0x92 fp=0xc438085c78 sp=0xc438085c28
net/http.HandlerFunc.ServeHTTP(0x893c08, 0xc3d3e0, 0xc4287aa410, 0xc4336d50e0)
/usr/local/go/src/net/http/server.go:1726 +0x44 fp=0xc438085ca0 sp=0xc438085c78
net/http.(*ServeMux).ServeHTTP(0xc53c40, 0xc3d3e0, 0xc4287aa410, 0xc4336d50e0)
/usr/local/go/src/net/http/server.go:2022 +0x7f fp=0xc438085ce0 sp=0xc438085ca0
net/http.serverHandler.ServeHTTP(0xc420138100, 0xc3d3e0, 0xc4287aa410, 0xc4336d50e0)
/usr/local/go/src/net/http/server.go:2202 +0x7d fp=0xc438085d28 sp=0xc438085ce0
net/http.(*conn).serve(0xc42badfc80, 0xc3de60, 0xc45161aec0)
/usr/local/go/src/net/http/server.go:1579 +0x4b7 fp=0xc438085f88 sp=0xc438085d28
runtime.goexit()
/usr/local/go/src/runtime/asm_amd64.s:2086 +0x1 fp=0xc438085f90 sp=0xc438085f88
created by net/http.(*Server).Serve
/usr/local/go/src/net/http/server.go:2293 +0x44d

我是小白，不知道为什么graph启动报：IOWorkerNum must be 2^N, current IOWorkerNum is 0这个错，请大神帮助解答一下。

刚刚开始试试open-falcon，安装还顺利 正在熟悉，发现graph两次自动down掉了。下面的报错是什么原因呢？

2016/03/24 23:24:18 cfg.go:90: g.ParseConfig ok, file cfg.json
2016/03/24 23:24:18 db.go:26: g.InitDB ok
2016/03/24 23:24:18 rrdtool.go:55: rrdtool.Start ok
2016/03/24 23:24:18 index.go:15: index.Start ok
2016/03/24 23:24:18 main.go:20: 15851 register signal notify
2016/03/24 23:24:18 rpc.go:54: rpc.Start ok, listening on 0.0.0.0:6070
2016/03/24 23:24:18 http.go:96: http listening 0.0.0.0:6071
panic: runtime error: cgo argument has Go pointer to Go pointer

goroutine 14 [running]:
panic(0x90f2e0, 0xc820315e30)
/usr/local/go/src/runtime/panic.go:464 +0x3e6
github.com/open-falcon/rrdlite.(_Updater).update(0xc820036de8, 0xc82031c240, 0x6, 0x8, 0x0, 0x0)
/workspace/golang/src/github.com/open-falcon/rrdlite/rrd_c.go:70 +0x100
github.com/open-falcon/rrdlite.(_Updater).Update(0xc820036de8, 0x0, 0x0, 0x0, 0x0, 0x0)
/workspace/golang/src/github.com/open-falcon/rrdlite/rrd.go:130 +0x27f
github.com/open-falcon/graph/rrdtool.update(0xc8202ec4b0, 0x45, 0xc82031a210, 0x6, 0x6, 0x0, 0x0)
/workspace/golang/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:117 +0x45e
github.com/open-falcon/graph/rrdtool.flushrrd(0xc8202ec4b0, 0x45, 0xc82031a210, 0x6, 0x6, 0x0, 0x0)
/workspace/golang/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:142 +0x1a2
github.com/open-falcon/graph/rrdtool.ioWorker()
/workspace/golang/src/github.com/open-falcon/graph/rrdtool/sync_disk.go:95 +0x38f
created by github.com/open-falcon/graph/rrdtool.Start
/workspace/golang/src/github.com/open-falcon/graph/rrdtool/rrdtool.go:54 +0x24f

代码http/debug_http.go中 141行
if item.Step < g.MIN_STEP {
item.Step = g.MIN_STEP
}
为什么步长小于30s的强制设定成30s呢，而且是写死没法配置，不太理解，这个设计是出于什么考虑

RT

现在graph和mysql连接后，是否对wait_timeout的心跳检测，今天在日志中发现了如下日志：
[MySQL] 2015/09/14 14:52:41 packets.go:118: write tcp 127.0.0.1:3306: broken pipe
[MySQL] 2015/09/14 14:52:41 packets.go:118: write tcp 127.0.0.1:3306: broken pipe
[MySQL] 2015/09/14 14:52:41 packets.go:118: write tcp 127.0.0.1:3306: broken pipe
[MySQL] 2015/09/14 14:52:41 packets.go:118: write tcp 127.0.0.1:3306: broken pipe
如果有该如何配置呢？

2015/07/16 17:09:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:10:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:10:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:11:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:11:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:12:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:12:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:13:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:13:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:14:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:14:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist
2015/07/16 17:15:00 index_update_incr_task.go:90: 127.0.0.1 Error 1146: Table 'graph.endpoint' doesn't exist

“我们参考RRDtool的理念，在数据每次存入的时候，会自动进行采样、归档。在默认的归档策略，一分钟push一次的频率下，历史数据保存5年。同时为了不丢失信息量，数据归档的时候，会按照平均值采样、最大值采样、最小值采样存三份。用户在查询某个metric，在过去一年的历史数据时，Graph会选择最合适的采样频率，返回采样过后的数据，提高了数据查询速度。”

对这段话，有几个疑问：
1、数据每次存入，怎么理解？什么情况下会触发一次数据存储？一次的存入量，怎么衡量？
2、一分钟push一次的频率下，历史数据保存5年。频率降低的话，是否可以存储更久的数据？
3、对于较久远的历史数据，保存的数据点会不会减少？

手动上报的数据，用的 book 里面那个例子

如果只上报一次，Dashboard里出现的那个单点有约一半的概率会在接下来的20min内消失，去 graph/history 里查也显示没有数据

请问这是怎么回事？

PS：连续上报的数据没问题。

`nowTs := time.Now().Unix()
lastUpTs := nowTs - nowTs%int64(step)
rra1StartTs := lastUpTs - int64(rrdtool.RRA1PointCnt*step)

// consolidated, do not merge
if start_ts < rra1StartTs {
	resp.Values = datas
	goto _RETURN_OK
}`

rra1StartTs 代表什么意思？怎样理解这段代码？