You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

mountstats.go 18KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616
  1. // Copyright 2018 The Prometheus Authors
  2. // Licensed under the Apache License, Version 2.0 (the "License");
  3. // you may not use this file except in compliance with the License.
  4. // You may obtain a copy of the License at
  5. //
  6. // http://www.apache.org/licenses/LICENSE-2.0
  7. //
  8. // Unless required by applicable law or agreed to in writing, software
  9. // distributed under the License is distributed on an "AS IS" BASIS,
  10. // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  11. // See the License for the specific language governing permissions and
  12. // limitations under the License.
  13. package procfs
  14. // While implementing parsing of /proc/[pid]/mountstats, this blog was used
  15. // heavily as a reference:
  16. // https://utcc.utoronto.ca/~cks/space/blog/linux/NFSMountstatsIndex
  17. //
  18. // Special thanks to Chris Siebenmann for all of his posts explaining the
  19. // various statistics available for NFS.
  20. import (
  21. "bufio"
  22. "fmt"
  23. "io"
  24. "strconv"
  25. "strings"
  26. "time"
  27. )
  28. // Constants shared between multiple functions.
  29. const (
  30. deviceEntryLen = 8
  31. fieldBytesLen = 8
  32. fieldEventsLen = 27
  33. statVersion10 = "1.0"
  34. statVersion11 = "1.1"
  35. fieldTransport10TCPLen = 10
  36. fieldTransport10UDPLen = 7
  37. fieldTransport11TCPLen = 13
  38. fieldTransport11UDPLen = 10
  39. )
  40. // A Mount is a device mount parsed from /proc/[pid]/mountstats.
  41. type Mount struct {
  42. // Name of the device.
  43. Device string
  44. // The mount point of the device.
  45. Mount string
  46. // The filesystem type used by the device.
  47. Type string
  48. // If available additional statistics related to this Mount.
  49. // Use a type assertion to determine if additional statistics are available.
  50. Stats MountStats
  51. }
  52. // A MountStats is a type which contains detailed statistics for a specific
  53. // type of Mount.
  54. type MountStats interface {
  55. mountStats()
  56. }
  57. // A MountStatsNFS is a MountStats implementation for NFSv3 and v4 mounts.
  58. type MountStatsNFS struct {
  59. // The version of statistics provided.
  60. StatVersion string
  61. // The optional mountaddr of the NFS mount.
  62. MountAddress string
  63. // The age of the NFS mount.
  64. Age time.Duration
  65. // Statistics related to byte counters for various operations.
  66. Bytes NFSBytesStats
  67. // Statistics related to various NFS event occurrences.
  68. Events NFSEventsStats
  69. // Statistics broken down by filesystem operation.
  70. Operations []NFSOperationStats
  71. // Statistics about the NFS RPC transport.
  72. Transport NFSTransportStats
  73. }
  74. // mountStats implements MountStats.
  75. func (m MountStatsNFS) mountStats() {}
  76. // A NFSBytesStats contains statistics about the number of bytes read and written
  77. // by an NFS client to and from an NFS server.
  78. type NFSBytesStats struct {
  79. // Number of bytes read using the read() syscall.
  80. Read uint64
  81. // Number of bytes written using the write() syscall.
  82. Write uint64
  83. // Number of bytes read using the read() syscall in O_DIRECT mode.
  84. DirectRead uint64
  85. // Number of bytes written using the write() syscall in O_DIRECT mode.
  86. DirectWrite uint64
  87. // Number of bytes read from the NFS server, in total.
  88. ReadTotal uint64
  89. // Number of bytes written to the NFS server, in total.
  90. WriteTotal uint64
  91. // Number of pages read directly via mmap()'d files.
  92. ReadPages uint64
  93. // Number of pages written directly via mmap()'d files.
  94. WritePages uint64
  95. }
  96. // A NFSEventsStats contains statistics about NFS event occurrences.
  97. type NFSEventsStats struct {
  98. // Number of times cached inode attributes are re-validated from the server.
  99. InodeRevalidate uint64
  100. // Number of times cached dentry nodes are re-validated from the server.
  101. DnodeRevalidate uint64
  102. // Number of times an inode cache is cleared.
  103. DataInvalidate uint64
  104. // Number of times cached inode attributes are invalidated.
  105. AttributeInvalidate uint64
  106. // Number of times files or directories have been open()'d.
  107. VFSOpen uint64
  108. // Number of times a directory lookup has occurred.
  109. VFSLookup uint64
  110. // Number of times permissions have been checked.
  111. VFSAccess uint64
  112. // Number of updates (and potential writes) to pages.
  113. VFSUpdatePage uint64
  114. // Number of pages read directly via mmap()'d files.
  115. VFSReadPage uint64
  116. // Number of times a group of pages have been read.
  117. VFSReadPages uint64
  118. // Number of pages written directly via mmap()'d files.
  119. VFSWritePage uint64
  120. // Number of times a group of pages have been written.
  121. VFSWritePages uint64
  122. // Number of times directory entries have been read with getdents().
  123. VFSGetdents uint64
  124. // Number of times attributes have been set on inodes.
  125. VFSSetattr uint64
  126. // Number of pending writes that have been forcefully flushed to the server.
  127. VFSFlush uint64
  128. // Number of times fsync() has been called on directories and files.
  129. VFSFsync uint64
  130. // Number of times locking has been attempted on a file.
  131. VFSLock uint64
  132. // Number of times files have been closed and released.
  133. VFSFileRelease uint64
  134. // Unknown. Possibly unused.
  135. CongestionWait uint64
  136. // Number of times files have been truncated.
  137. Truncation uint64
  138. // Number of times a file has been grown due to writes beyond its existing end.
  139. WriteExtension uint64
  140. // Number of times a file was removed while still open by another process.
  141. SillyRename uint64
  142. // Number of times the NFS server gave less data than expected while reading.
  143. ShortRead uint64
  144. // Number of times the NFS server wrote less data than expected while writing.
  145. ShortWrite uint64
  146. // Number of times the NFS server indicated EJUKEBOX; retrieving data from
  147. // offline storage.
  148. JukeboxDelay uint64
  149. // Number of NFS v4.1+ pNFS reads.
  150. PNFSRead uint64
  151. // Number of NFS v4.1+ pNFS writes.
  152. PNFSWrite uint64
  153. }
  154. // A NFSOperationStats contains statistics for a single operation.
  155. type NFSOperationStats struct {
  156. // The name of the operation.
  157. Operation string
  158. // Number of requests performed for this operation.
  159. Requests uint64
  160. // Number of times an actual RPC request has been transmitted for this operation.
  161. Transmissions uint64
  162. // Number of times a request has had a major timeout.
  163. MajorTimeouts uint64
  164. // Number of bytes sent for this operation, including RPC headers and payload.
  165. BytesSent uint64
  166. // Number of bytes received for this operation, including RPC headers and payload.
  167. BytesReceived uint64
  168. // Duration all requests spent queued for transmission before they were sent.
  169. CumulativeQueueTime time.Duration
  170. // Duration it took to get a reply back after the request was transmitted.
  171. CumulativeTotalResponseTime time.Duration
  172. // Duration from when a request was enqueued to when it was completely handled.
  173. CumulativeTotalRequestTime time.Duration
  174. }
  175. // A NFSTransportStats contains statistics for the NFS mount RPC requests and
  176. // responses.
  177. type NFSTransportStats struct {
  178. // The transport protocol used for the NFS mount.
  179. Protocol string
  180. // The local port used for the NFS mount.
  181. Port uint64
  182. // Number of times the client has had to establish a connection from scratch
  183. // to the NFS server.
  184. Bind uint64
  185. // Number of times the client has made a TCP connection to the NFS server.
  186. Connect uint64
  187. // Duration (in jiffies, a kernel internal unit of time) the NFS mount has
  188. // spent waiting for connections to the server to be established.
  189. ConnectIdleTime uint64
  190. // Duration since the NFS mount last saw any RPC traffic.
  191. IdleTime time.Duration
  192. // Number of RPC requests for this mount sent to the NFS server.
  193. Sends uint64
  194. // Number of RPC responses for this mount received from the NFS server.
  195. Receives uint64
  196. // Number of times the NFS server sent a response with a transaction ID
  197. // unknown to this client.
  198. BadTransactionIDs uint64
  199. // A running counter, incremented on each request as the current difference
  200. // ebetween sends and receives.
  201. CumulativeActiveRequests uint64
  202. // A running counter, incremented on each request by the current backlog
  203. // queue size.
  204. CumulativeBacklog uint64
  205. // Stats below only available with stat version 1.1.
  206. // Maximum number of simultaneously active RPC requests ever used.
  207. MaximumRPCSlotsUsed uint64
  208. // A running counter, incremented on each request as the current size of the
  209. // sending queue.
  210. CumulativeSendingQueue uint64
  211. // A running counter, incremented on each request as the current size of the
  212. // pending queue.
  213. CumulativePendingQueue uint64
  214. }
  215. // parseMountStats parses a /proc/[pid]/mountstats file and returns a slice
  216. // of Mount structures containing detailed information about each mount.
  217. // If available, statistics for each mount are parsed as well.
  218. func parseMountStats(r io.Reader) ([]*Mount, error) {
  219. const (
  220. device = "device"
  221. statVersionPrefix = "statvers="
  222. nfs3Type = "nfs"
  223. nfs4Type = "nfs4"
  224. )
  225. var mounts []*Mount
  226. s := bufio.NewScanner(r)
  227. for s.Scan() {
  228. // Only look for device entries in this function
  229. ss := strings.Fields(string(s.Bytes()))
  230. if len(ss) == 0 || ss[0] != device {
  231. continue
  232. }
  233. m, err := parseMount(ss)
  234. if err != nil {
  235. return nil, err
  236. }
  237. // Does this mount also possess statistics information?
  238. if len(ss) > deviceEntryLen {
  239. // Only NFSv3 and v4 are supported for parsing statistics
  240. if m.Type != nfs3Type && m.Type != nfs4Type {
  241. return nil, fmt.Errorf("cannot parse MountStats for fstype %q", m.Type)
  242. }
  243. statVersion := strings.TrimPrefix(ss[8], statVersionPrefix)
  244. stats, err := parseMountStatsNFS(s, statVersion)
  245. if err != nil {
  246. return nil, err
  247. }
  248. m.Stats = stats
  249. }
  250. mounts = append(mounts, m)
  251. }
  252. return mounts, s.Err()
  253. }
  254. // parseMount parses an entry in /proc/[pid]/mountstats in the format:
  255. // device [device] mounted on [mount] with fstype [type]
  256. func parseMount(ss []string) (*Mount, error) {
  257. if len(ss) < deviceEntryLen {
  258. return nil, fmt.Errorf("invalid device entry: %v", ss)
  259. }
  260. // Check for specific words appearing at specific indices to ensure
  261. // the format is consistent with what we expect
  262. format := []struct {
  263. i int
  264. s string
  265. }{
  266. {i: 0, s: "device"},
  267. {i: 2, s: "mounted"},
  268. {i: 3, s: "on"},
  269. {i: 5, s: "with"},
  270. {i: 6, s: "fstype"},
  271. }
  272. for _, f := range format {
  273. if ss[f.i] != f.s {
  274. return nil, fmt.Errorf("invalid device entry: %v", ss)
  275. }
  276. }
  277. return &Mount{
  278. Device: ss[1],
  279. Mount: ss[4],
  280. Type: ss[7],
  281. }, nil
  282. }
  283. // parseMountStatsNFS parses a MountStatsNFS by scanning additional information
  284. // related to NFS statistics.
  285. func parseMountStatsNFS(s *bufio.Scanner, statVersion string) (*MountStatsNFS, error) {
  286. // Field indicators for parsing specific types of data
  287. const (
  288. fieldOpts = "opts:"
  289. fieldAge = "age:"
  290. fieldBytes = "bytes:"
  291. fieldEvents = "events:"
  292. fieldPerOpStats = "per-op"
  293. fieldTransport = "xprt:"
  294. )
  295. stats := &MountStatsNFS{
  296. StatVersion: statVersion,
  297. }
  298. for s.Scan() {
  299. ss := strings.Fields(string(s.Bytes()))
  300. if len(ss) == 0 {
  301. break
  302. }
  303. if len(ss) < 2 {
  304. return nil, fmt.Errorf("not enough information for NFS stats: %v", ss)
  305. }
  306. switch ss[0] {
  307. case fieldOpts:
  308. for _, opt := range strings.Split(ss[1], ",") {
  309. split := strings.Split(opt, "=")
  310. if len(split) == 2 && split[0] == "mountaddr" {
  311. stats.MountAddress = split[1]
  312. }
  313. }
  314. case fieldAge:
  315. // Age integer is in seconds
  316. d, err := time.ParseDuration(ss[1] + "s")
  317. if err != nil {
  318. return nil, err
  319. }
  320. stats.Age = d
  321. case fieldBytes:
  322. bstats, err := parseNFSBytesStats(ss[1:])
  323. if err != nil {
  324. return nil, err
  325. }
  326. stats.Bytes = *bstats
  327. case fieldEvents:
  328. estats, err := parseNFSEventsStats(ss[1:])
  329. if err != nil {
  330. return nil, err
  331. }
  332. stats.Events = *estats
  333. case fieldTransport:
  334. if len(ss) < 3 {
  335. return nil, fmt.Errorf("not enough information for NFS transport stats: %v", ss)
  336. }
  337. tstats, err := parseNFSTransportStats(ss[1:], statVersion)
  338. if err != nil {
  339. return nil, err
  340. }
  341. stats.Transport = *tstats
  342. }
  343. // When encountering "per-operation statistics", we must break this
  344. // loop and parse them separately to ensure we can terminate parsing
  345. // before reaching another device entry; hence why this 'if' statement
  346. // is not just another switch case
  347. if ss[0] == fieldPerOpStats {
  348. break
  349. }
  350. }
  351. if err := s.Err(); err != nil {
  352. return nil, err
  353. }
  354. // NFS per-operation stats appear last before the next device entry
  355. perOpStats, err := parseNFSOperationStats(s)
  356. if err != nil {
  357. return nil, err
  358. }
  359. stats.Operations = perOpStats
  360. return stats, nil
  361. }
  362. // parseNFSBytesStats parses a NFSBytesStats line using an input set of
  363. // integer fields.
  364. func parseNFSBytesStats(ss []string) (*NFSBytesStats, error) {
  365. if len(ss) != fieldBytesLen {
  366. return nil, fmt.Errorf("invalid NFS bytes stats: %v", ss)
  367. }
  368. ns := make([]uint64, 0, fieldBytesLen)
  369. for _, s := range ss {
  370. n, err := strconv.ParseUint(s, 10, 64)
  371. if err != nil {
  372. return nil, err
  373. }
  374. ns = append(ns, n)
  375. }
  376. return &NFSBytesStats{
  377. Read: ns[0],
  378. Write: ns[1],
  379. DirectRead: ns[2],
  380. DirectWrite: ns[3],
  381. ReadTotal: ns[4],
  382. WriteTotal: ns[5],
  383. ReadPages: ns[6],
  384. WritePages: ns[7],
  385. }, nil
  386. }
  387. // parseNFSEventsStats parses a NFSEventsStats line using an input set of
  388. // integer fields.
  389. func parseNFSEventsStats(ss []string) (*NFSEventsStats, error) {
  390. if len(ss) != fieldEventsLen {
  391. return nil, fmt.Errorf("invalid NFS events stats: %v", ss)
  392. }
  393. ns := make([]uint64, 0, fieldEventsLen)
  394. for _, s := range ss {
  395. n, err := strconv.ParseUint(s, 10, 64)
  396. if err != nil {
  397. return nil, err
  398. }
  399. ns = append(ns, n)
  400. }
  401. return &NFSEventsStats{
  402. InodeRevalidate: ns[0],
  403. DnodeRevalidate: ns[1],
  404. DataInvalidate: ns[2],
  405. AttributeInvalidate: ns[3],
  406. VFSOpen: ns[4],
  407. VFSLookup: ns[5],
  408. VFSAccess: ns[6],
  409. VFSUpdatePage: ns[7],
  410. VFSReadPage: ns[8],
  411. VFSReadPages: ns[9],
  412. VFSWritePage: ns[10],
  413. VFSWritePages: ns[11],
  414. VFSGetdents: ns[12],
  415. VFSSetattr: ns[13],
  416. VFSFlush: ns[14],
  417. VFSFsync: ns[15],
  418. VFSLock: ns[16],
  419. VFSFileRelease: ns[17],
  420. CongestionWait: ns[18],
  421. Truncation: ns[19],
  422. WriteExtension: ns[20],
  423. SillyRename: ns[21],
  424. ShortRead: ns[22],
  425. ShortWrite: ns[23],
  426. JukeboxDelay: ns[24],
  427. PNFSRead: ns[25],
  428. PNFSWrite: ns[26],
  429. }, nil
  430. }
  431. // parseNFSOperationStats parses a slice of NFSOperationStats by scanning
  432. // additional information about per-operation statistics until an empty
  433. // line is reached.
  434. func parseNFSOperationStats(s *bufio.Scanner) ([]NFSOperationStats, error) {
  435. const (
  436. // Number of expected fields in each per-operation statistics set
  437. numFields = 9
  438. )
  439. var ops []NFSOperationStats
  440. for s.Scan() {
  441. ss := strings.Fields(string(s.Bytes()))
  442. if len(ss) == 0 {
  443. // Must break when reading a blank line after per-operation stats to
  444. // enable top-level function to parse the next device entry
  445. break
  446. }
  447. if len(ss) != numFields {
  448. return nil, fmt.Errorf("invalid NFS per-operations stats: %v", ss)
  449. }
  450. // Skip string operation name for integers
  451. ns := make([]uint64, 0, numFields-1)
  452. for _, st := range ss[1:] {
  453. n, err := strconv.ParseUint(st, 10, 64)
  454. if err != nil {
  455. return nil, err
  456. }
  457. ns = append(ns, n)
  458. }
  459. ops = append(ops, NFSOperationStats{
  460. Operation: strings.TrimSuffix(ss[0], ":"),
  461. Requests: ns[0],
  462. Transmissions: ns[1],
  463. MajorTimeouts: ns[2],
  464. BytesSent: ns[3],
  465. BytesReceived: ns[4],
  466. CumulativeQueueTime: time.Duration(ns[5]) * time.Millisecond,
  467. CumulativeTotalResponseTime: time.Duration(ns[6]) * time.Millisecond,
  468. CumulativeTotalRequestTime: time.Duration(ns[7]) * time.Millisecond,
  469. })
  470. }
  471. return ops, s.Err()
  472. }
  473. // parseNFSTransportStats parses a NFSTransportStats line using an input set of
  474. // integer fields matched to a specific stats version.
  475. func parseNFSTransportStats(ss []string, statVersion string) (*NFSTransportStats, error) {
  476. // Extract the protocol field. It is the only string value in the line
  477. protocol := ss[0]
  478. ss = ss[1:]
  479. switch statVersion {
  480. case statVersion10:
  481. var expectedLength int
  482. if protocol == "tcp" {
  483. expectedLength = fieldTransport10TCPLen
  484. } else if protocol == "udp" {
  485. expectedLength = fieldTransport10UDPLen
  486. } else {
  487. return nil, fmt.Errorf("invalid NFS protocol \"%s\" in stats 1.0 statement: %v", protocol, ss)
  488. }
  489. if len(ss) != expectedLength {
  490. return nil, fmt.Errorf("invalid NFS transport stats 1.0 statement: %v", ss)
  491. }
  492. case statVersion11:
  493. var expectedLength int
  494. if protocol == "tcp" {
  495. expectedLength = fieldTransport11TCPLen
  496. } else if protocol == "udp" {
  497. expectedLength = fieldTransport11UDPLen
  498. } else {
  499. return nil, fmt.Errorf("invalid NFS protocol \"%s\" in stats 1.1 statement: %v", protocol, ss)
  500. }
  501. if len(ss) != expectedLength {
  502. return nil, fmt.Errorf("invalid NFS transport stats 1.1 statement: %v", ss)
  503. }
  504. default:
  505. return nil, fmt.Errorf("unrecognized NFS transport stats version: %q", statVersion)
  506. }
  507. // Allocate enough for v1.1 stats since zero value for v1.1 stats will be okay
  508. // in a v1.0 response. Since the stat length is bigger for TCP stats, we use
  509. // the TCP length here.
  510. //
  511. // Note: slice length must be set to length of v1.1 stats to avoid a panic when
  512. // only v1.0 stats are present.
  513. // See: https://github.com/prometheus/node_exporter/issues/571.
  514. ns := make([]uint64, fieldTransport11TCPLen)
  515. for i, s := range ss {
  516. n, err := strconv.ParseUint(s, 10, 64)
  517. if err != nil {
  518. return nil, err
  519. }
  520. ns[i] = n
  521. }
  522. // The fields differ depending on the transport protocol (TCP or UDP)
  523. // From https://utcc.utoronto.ca/%7Ecks/space/blog/linux/NFSMountstatsXprt
  524. //
  525. // For the udp RPC transport there is no connection count, connect idle time,
  526. // or idle time (fields #3, #4, and #5); all other fields are the same. So
  527. // we set them to 0 here.
  528. if protocol == "udp" {
  529. ns = append(ns[:2], append(make([]uint64, 3), ns[2:]...)...)
  530. }
  531. return &NFSTransportStats{
  532. Protocol: protocol,
  533. Port: ns[0],
  534. Bind: ns[1],
  535. Connect: ns[2],
  536. ConnectIdleTime: ns[3],
  537. IdleTime: time.Duration(ns[4]) * time.Second,
  538. Sends: ns[5],
  539. Receives: ns[6],
  540. BadTransactionIDs: ns[7],
  541. CumulativeActiveRequests: ns[8],
  542. CumulativeBacklog: ns[9],
  543. MaximumRPCSlotsUsed: ns[10],
  544. CumulativeSendingQueue: ns[11],
  545. CumulativePendingQueue: ns[12],
  546. }, nil
  547. }