com.bigdata.counters.query.CounterSetQuery Maven / Gradle / Ivy
/*
Copyright (C) SYSTAP, LLC DBA Blazegraph 2006-2016. All rights reserved.
Contact:
SYSTAP, LLC DBA Blazegraph
2501 Calvert ST NW #106
Washington, DC 20008
[email protected]
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* Created on Apr 6, 2009
*/
package com.bigdata.counters.query;
import java.io.BufferedReader;
import java.io.BufferedWriter;
import java.io.File;
import java.io.FileFilter;
import java.io.FileNotFoundException;
import java.io.FileReader;
import java.io.FileWriter;
import java.io.IOException;
import java.io.Writer;
import java.net.URL;
import java.util.ArrayList;
import java.util.Collection;
import java.util.LinkedList;
import java.util.List;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;
import java.util.regex.Pattern;
import javax.xml.parsers.ParserConfigurationException;
import org.apache.log4j.Logger;
import org.xml.sax.SAXException;
import com.bigdata.Banner;
import com.bigdata.counters.CounterSet;
import com.bigdata.counters.IHostCounters;
import com.bigdata.counters.IRequiredHostCounters;
import com.bigdata.counters.PeriodEnum;
import com.bigdata.counters.XMLUtility;
import com.bigdata.counters.httpd.DummyEventReportingService;
import com.bigdata.counters.render.IRenderer;
import com.bigdata.counters.render.RendererFactory;
import com.bigdata.service.Event;
import com.bigdata.util.httpd.NanoHTTPD;
/**
* Utility to extract a batch of performance counters from a collection of
* logged XML counter set files. This utility accepts file(s) giving the URLs
* which would be used to demand the corresponding performance counters against
* the live bigdata federation. The URLs listed in that file are parsed. The
* host and port information are ignored, but the URL query parameters are
* extracted and used to configured a set of {@link URLQueryModel}s.
*
* A single pass is made through the specified XML counter set files. Each file
* is read into memory by itself, and each query implied by a listed URL is run
* against the in-memory {@link CounterSet} hierarchy. The results are collected
* in independent {@link CounterSet} provisioned for the specified reporting
* units, etc. Once the last XML counter set file has been processed, the
* various reports requested by the listed URLs are generated.
*
* For each generated report, the name of the file on which the report will be
* written is taken from the name of the counter whose value was extracted for
* that report. This filename may be overridden by including the URL query
* parameter {@value URLQueryModel#FILE}, which specifies the file on which to
* write the report for that query.
*
* @author Bryan Thompson
* @version $Id$
*
* @see IHostCounters
* @see IRequiredHostCounters
*
* @todo When rendering HTML output using flot, the flot resources need to be
* available in order to view the graphs. They should be written once into
* the output directory and the links in the (X)HTML output should resolve
* them there.
*
* @todo Permit nsamples to be specified in units of minutes, hours, days. E.g.,
* 3d would be 3 days.
*
* @todo Performance for long runs could be improved if we use more efficient
* classes for mutable strings in {@link XMLUtility} and perhaps
* {@link CounterSet}. E.g., mg4j mutable string or
* http://javolution.org/.
*/
public class CounterSetQuery {
static private final Logger log = Logger.getLogger(CounterSetQuery.class);
/**
* Reads a list of {@link URL}s from a file. Blank lines and comment lines
* are ignored.
*
* @param file
* A file containing URLs, blank lines, and comment lines (which
* start with '#').
*
* @return A list of the URLs read from the file.
*
* @throws IOException
*/
private static Collection readURLsFromFile(final File file) throws IOException {
if(log.isInfoEnabled())
log.info("Reading queries: "+file);
final List tmp = new LinkedList();
final BufferedReader r = new BufferedReader(new FileReader(file));
try {
String s;
while ((s = r.readLine()) != null) {
s = s.trim();
if (s.isEmpty())
continue;
if (s.startsWith("#"))
continue;
tmp.add(new URL(s));
}
} finally {
r.close();
}
return tmp;
}
/**
* Reads URLs from a file or all files (recursively) in a directory.
*
* @param file
* The file or directory.
*
* @return The URLs read from the file(s).
*
* @throws IOException
*/
static private Collection readURLs(final File file) throws IOException {
/*
* note: duplicates are not filtered out but this preserves the
* evaluation order.
*/
final Collection urls = new LinkedList();
if (file.isDirectory()) {
final File[] files = file.listFiles();
for(File f : files) {
if (f.isHidden())
continue;
urls.addAll(readURLsFromFile(f));
}
} else {
urls.addAll(readURLsFromFile(file));
}
return urls;
}
private static void readFiles(final Collection counterSetFiles,
final CounterSet counterSet, final int nsamples,
final PeriodEnum period, final Pattern regex) throws IOException,
SAXException, ParserConfigurationException, InterruptedException,
ExecutionException {
// flatten directories in the list of files.
final Collection flatFileList = QueryUtil.collectFiles(counterSetFiles,
new FileFilter() {
public boolean accept(File pathname) {
return !pathname.isHidden()
&& pathname.getName().endsWith(".xml");
}
});
if (log.isInfoEnabled())
log.info("Reading performance counters from "
+ flatFileList.size() + " sources.");
// read the files
if (false/* sequential */) {
// process the files one at a time.
readFilesSequential(flatFileList, counterSet, nsamples, period,
regex);
} else {
// process the files in parallel.
readFilesParallel(flatFileList, counterSet, nsamples, period,
regex);
}
}
private static void readFilesSequential(
final Collection counterSetFiles,
final CounterSet counterSet, final int nsamples,
final PeriodEnum period, final Pattern regex) throws IOException,
SAXException, ParserConfigurationException {
for (File file : counterSetFiles) {
if(log.isInfoEnabled())
log.info("Reading file: " + file);
QueryUtil.readCountersFromFile(file, counterSet, regex, nsamples,
period);
}
}
private static void readFilesParallel(
final Collection counterSetFiles,
final CounterSet counterSet, final int nsamples,
final PeriodEnum period, final Pattern regex) throws IOException,
SAXException, ParserConfigurationException, InterruptedException, ExecutionException {
final int nfiles = counterSetFiles.size();
final List> tasks = new ArrayList>(nfiles);
for (File file : counterSetFiles) {
tasks.add(new QueryUtil.ReadCounterSetXMLFileTask(file, counterSet,
nsamples, period, regex));
}
final ExecutorService service = Executors.newFixedThreadPool(nfiles);
final List> futures;
try {
// run all tasks.
futures = service.invokeAll(tasks);
} finally {
service.shutdownNow();
}
int i = 0;
int nerrors = 0;
for(Future future : futures) {
// look for errors in the tasks.
try {
future.get();
} catch(ExecutionException ex) {
if(ex.getCause() instanceof SAXException) {
/*
* Sometimes you can get a partial XML file if the LBS was
* in the process of generating the file when it was copied.
* This shows up as a SAXException. Rather than dying, this
* just logs a warning and continues.
*/
log.warn("Could not parse file (ignored): " + tasks.get(i), ex);
nerrors++;
continue;
}
}
i++;
}
if (nerrors != 0)
log.error("There were " + nerrors + " errors.");
}
/**
* Utility class for running extracting data from performance counter dumps
* and running various kinds of reports on those data.
*
* Usage:
*
* - -outputDir
* - The output directory (default is the current working directory).
* - -mimeType
* - The default MIME type for the rendered reports. The default is
*
text/plain, but can be overridden on a query by query basis
* using {@link URLQueryModel#MIMETYPE}.
* - -nsamples
* - Override for the default #of history samples to be retained. It is an
* error if there are more distinct samples in the processed XML counter set
* files (that is, if the #of time periods sampled exceeds this many
* samples). If there are fewer, then some internal arrays will be
* dimensioned larger than is otherwise necessary.
* - -events <file>
* - A file containing tab-delimited {@link Event}s. The {@link Event}s
* are not required for simple performance counter views.
* - -queries <file>
* - A file, or directory of files, containing a list of URLs, each of
* which is interpreted as a {@link URLQueryModel}.
* - <file>(s)
* - One or more XML counter set files or directories containing such
* files. All such files will be processed before the reports are generated.
*
*
*
* @param args
* Command line arguments.
*
* @throws IOException
* @throws ParserConfigurationException
* @throws SAXException
* @throws ExecutionException
* @throws InterruptedException
*/
public static void main(final String[] args) throws IOException,
SAXException, ParserConfigurationException, InterruptedException,
ExecutionException {
Banner.banner();
if (args.length == 0) {
System.err.println("See javadoc for usage.");
System.exit(1);
}
/*
* The events read from the file(s).
*/
final DummyEventReportingService service = new DummyEventReportingService();
// The default output format (text, html, etc.)
String defaultMimeType = NanoHTTPD.MIME_TEXT_PLAIN;
/*
* The #of slots to allocate (one slot per period of data to be read).
*
* Note: The default is 7 days of data if period is minutes.
*/
int nsamples = 60 * 24 * 7;
// The output directory defaults to the current working directory.
File outputDir = new File(".");
// the set of queries to be processed.
final List queries = new LinkedList();
// the set of counter set XML files to be processed.
final List counterSetFiles = new LinkedList();
for (int i = 0; i < args.length; i++) {
final String arg = args[i];
if (arg.startsWith("-")) {
if (arg.equals("-outputDir")) {
outputDir = new File(args[++i]);
if (log.isInfoEnabled()) {
log.info("outputDir: " + outputDir);
}
if(!outputDir.exists()) {
outputDir.mkdirs();
}
} else if (arg.equals("-mimeType")) {
defaultMimeType = args[++i];
} else if (arg.equals("-nsamples")) {
nsamples = Integer.valueOf(args[++i]);
if (nsamples <= 0)
throw new IllegalArgumentException(
"nslots must be positive.");
} else if (arg.equals("-events")) {
// @todo read list of event files once all args are parsed.
QueryUtil.readEvents(service, new File(args[++i]));
} else if (arg.equals("-queries")) {
final File file = new File(args[++i]);
final Collection urls = readURLs(file);
for (URL url : urls) {
queries.add(URLQueryModel.getInstance(url));
}
} else {
System.err.println("Unknown option: " + arg);
System.exit(1);
}
} else {
final File file = new File(arg);
if (!file.exists())
throw new FileNotFoundException(file.toString());
counterSetFiles.add(file);
}
}
if (queries.isEmpty()) {
throw new RuntimeException("No queries were specified.");
}
if (counterSetFiles.isEmpty()) {
throw new RuntimeException("No counter set files were specified.");
}
/*
* Compute a regular expression which will match anything which would
* have been matched by the individual URLs. E.g., the OR of the
* individual regular expressions entailed by each URL when interpreted
* as a query.
*/
final Pattern regex;
{
final List tmp = new LinkedList();
for (URLQueryModel model : queries) {
if (model.pattern != null) {
tmp.add(model.pattern);
}
}
regex = QueryUtil.getPattern(tmp);
}
/*
* Read counters accepted by the optional filter into the counter set to
* be served.
*
* @todo this does not support reading at different periods for each
* query.
*/
// The performance counters read from the file(s).
final CounterSet counterSet = new CounterSet();
readFiles(counterSetFiles, counterSet, nsamples, PeriodEnum.Minutes,
regex);
/*
* Run each query in turn against the filtered pre-loaded counter set.
*/
if (log.isInfoEnabled())
log.info("Evaluating " + queries.size() + " queries.");
for (URLQueryModel model : queries) {
try {
final IRenderer renderer = RendererFactory.get(model,
new CounterSetSelector(counterSet), defaultMimeType);
/*
* Render on a file. The file can be specified by a URL query
* parameter.
*
* @todo Use the munged counter path / counter name (when one
* can be identified) as the default filename.
*/
File file;
if (model.file == null) {
file = File.createTempFile("query", ".out", outputDir);
} else {
if (!model.file.isAbsolute()) {
file = new File(outputDir, model.file.toString());
} else {
file = model.file;
}
}
if (file.getParentFile() != null
&& !file.getParentFile().exists()) {
if (log.isInfoEnabled()) {
log.info("Creating directory: " + file.getParentFile());
}
// make sure the parent directory exists.
file.getParentFile().mkdirs();
}
if (log.isInfoEnabled()) {
log.info("Writing file: " + file + " for query: "
+ model.uri);
}
final Writer w = new BufferedWriter(
new FileWriter(file, false/* append */));
try {
renderer.render(w);
w.flush();
} finally {
w.close();
}
} catch (Throwable t) {
log.error("Could not run query: " + model.uri, t);
}
}
}
}