From: Brian Warner <warner@allmydata.com>
Date: Thu, 24 Apr 2008 18:28:35 +0000 (-0700)
Subject: munin/tahoe_estimate_files.py: tool to estimate the total number of slots (mutable... 
X-Git-Tag: allmydata-tahoe-1.1.0~182
X-Git-Url: https://git.rkrishnan.org/components/com_hotproperty/simplejson/encoder.py.html?a=commitdiff_plain;h=d2a2b8a4ae099d8e44d290b5c856d145f91825c9;p=tahoe-lafs%2Ftahoe-lafs.git

munin/tahoe_estimate_files.py: tool to estimate the total number of slots (mutable and immutable combined) in the grid, from a small sample
---

diff --git a/misc/munin/tahoe_estimate_files.py b/misc/munin/tahoe_estimate_files.py
new file mode 100644
index 00000000..4723cdad
--- /dev/null
+++ b/misc/munin/tahoe_estimate_files.py
@@ -0,0 +1,49 @@
+#! /usr/bin/python
+
+import sys, os.path
+
+if len(sys.argv) > 1 and sys.argv[1] == "config":
+    print """\
+graph_title Tahoe File Estimate
+graph_vlabel files
+graph_category tahoe
+graph_info This graph shows the estimated number of files and directories present in the grid
+files.label files
+files.draw LINE2"""
+    sys.exit(0)
+
+# Edit this to point at some subset of storage directories.
+node_dirs = [os.path.expanduser("~amduser/prodnet/storage1"),
+             os.path.expanduser("~amduser/prodnet/storage2"),
+             os.path.expanduser("~amduser/prodnet/storage3"),
+             os.path.expanduser("~amduser/prodnet/storage4"),
+             ]
+
+sections = ["aa", "ab", "ac", "ad", "ae", "af", "ag", "ah", "ai", "aj"]
+# and edit this to reflect your default encoding's "total_shares" value, and
+# the total number of servers.
+N = 10
+num_servers = 20
+
+index_strings = set()
+for base in node_dirs:
+    for section in sections:
+        sampledir = os.path.join(base, "storage", "shares", section)
+        indices = os.listdir(sampledir)
+        index_strings.update(indices)
+unique_strings = len(index_strings)
+
+# the chance that any given file appears on any given server
+chance = 1.0 * N / num_servers
+
+# the chance that the file does *not* appear on the servers that we're
+# examining
+no_chance = (1-chance) ** len(node_dirs)
+
+# if a file has a 25% chance of not appearing in our sample, then we need to
+# raise our estimate by (1.25/1)
+correction = 1+no_chance
+#print "correction", correction
+
+files = unique_strings * (32*32/len(sections)) * correction
+print "files.value %d" % int(files)