khenaidoo | ab1f7bd | 2019-11-14 14:00:27 -0500 | [diff] [blame] | 1 | // Copyright 2019 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 | |
| 14 | package procfs |
| 15 | |
| 16 | // The PSI / pressure interface is described at |
| 17 | // https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/accounting/psi.txt |
| 18 | // Each resource (cpu, io, memory, ...) is exposed as a single file. |
| 19 | // Each file may contain up to two lines, one for "some" pressure and one for "full" pressure. |
| 20 | // Each line contains several averages (over n seconds) and a total in µs. |
| 21 | // |
| 22 | // Example io pressure file: |
| 23 | // > some avg10=0.06 avg60=0.21 avg300=0.99 total=8537362 |
| 24 | // > full avg10=0.00 avg60=0.13 avg300=0.96 total=8183134 |
| 25 | |
| 26 | import ( |
| 27 | "fmt" |
| 28 | "io" |
| 29 | "io/ioutil" |
| 30 | "os" |
| 31 | "strings" |
| 32 | ) |
| 33 | |
| 34 | const lineFormat = "avg10=%f avg60=%f avg300=%f total=%d" |
| 35 | |
| 36 | // PSILine is a single line of values as returned by /proc/pressure/* |
| 37 | // The Avg entries are averages over n seconds, as a percentage |
| 38 | // The Total line is in microseconds |
| 39 | type PSILine struct { |
| 40 | Avg10 float64 |
| 41 | Avg60 float64 |
| 42 | Avg300 float64 |
| 43 | Total uint64 |
| 44 | } |
| 45 | |
| 46 | // PSIStats represent pressure stall information from /proc/pressure/* |
| 47 | // Some indicates the share of time in which at least some tasks are stalled |
| 48 | // Full indicates the share of time in which all non-idle tasks are stalled simultaneously |
| 49 | type PSIStats struct { |
| 50 | Some *PSILine |
| 51 | Full *PSILine |
| 52 | } |
| 53 | |
| 54 | // PSIStatsForResource reads pressure stall information for the specified |
| 55 | // resource from /proc/pressure/<resource>. At time of writing this can be |
| 56 | // either "cpu", "memory" or "io". |
| 57 | func (fs FS) PSIStatsForResource(resource string) (PSIStats, error) { |
| 58 | file, err := os.Open(fs.proc.Path(fmt.Sprintf("%s/%s", "pressure", resource))) |
| 59 | if err != nil { |
| 60 | return PSIStats{}, fmt.Errorf("psi_stats: unavailable for %s", resource) |
| 61 | } |
| 62 | |
| 63 | defer file.Close() |
| 64 | return parsePSIStats(resource, file) |
| 65 | } |
| 66 | |
| 67 | // parsePSIStats parses the specified file for pressure stall information |
| 68 | func parsePSIStats(resource string, file io.Reader) (PSIStats, error) { |
| 69 | psiStats := PSIStats{} |
| 70 | stats, err := ioutil.ReadAll(file) |
| 71 | if err != nil { |
| 72 | return psiStats, fmt.Errorf("psi_stats: unable to read data for %s", resource) |
| 73 | } |
| 74 | |
| 75 | for _, l := range strings.Split(string(stats), "\n") { |
| 76 | prefix := strings.Split(l, " ")[0] |
| 77 | switch prefix { |
| 78 | case "some": |
| 79 | psi := PSILine{} |
| 80 | _, err := fmt.Sscanf(l, fmt.Sprintf("some %s", lineFormat), &psi.Avg10, &psi.Avg60, &psi.Avg300, &psi.Total) |
| 81 | if err != nil { |
| 82 | return PSIStats{}, err |
| 83 | } |
| 84 | psiStats.Some = &psi |
| 85 | case "full": |
| 86 | psi := PSILine{} |
| 87 | _, err := fmt.Sscanf(l, fmt.Sprintf("full %s", lineFormat), &psi.Avg10, &psi.Avg60, &psi.Avg300, &psi.Total) |
| 88 | if err != nil { |
| 89 | return PSIStats{}, err |
| 90 | } |
| 91 | psiStats.Full = &psi |
| 92 | default: |
| 93 | // If we encounter a line with an unknown prefix, ignore it and move on |
| 94 | // Should new measurement types be added in the future we'll simply ignore them instead |
| 95 | // of erroring on retrieval |
| 96 | continue |
| 97 | } |
| 98 | } |
| 99 | |
| 100 | return psiStats, nil |
| 101 | } |