Source code for girder

# -*- coding: utf-8 -*-
from pkg_resources import DistributionNotFound, get_distribution

    __version__ = get_distribution(__name__).version
except DistributionNotFound:
    # package is not installed
    __version__ = None

__license__ = 'Apache 2.0'

import cherrypy
import functools
import io
import logging
import logging.handlers
import os
import sys
import traceback

from girder.constants import LOG_ROOT, MAX_LOG_SIZE, LOG_BACKUP_COUNT, TerminalColor
from girder.utility import config, mkdir
from girder.utility._cache import cache, requestCache, rateLimitBuffer

_quiet = False
_originalStdOut = sys.stdout
_originalStdErr = sys.stderr
auditLogger = logging.getLogger('girder_audit')
logger = logging.getLogger('girder')
logger.setLevel(logging.DEBUG)  # Pass everything; let filters handle level-based filtering
config.loadConfig()  # Populate the config info at import time

[docs]class LogLevelFilter: """ Filter log records based on whether they are between a min and max level. """ def __init__(self, min, max): self.minLevel = min self.maxLevel = max def filter(self, logRecord): level = logRecord.levelno return self.maxLevel >= level >= self.minLevel
[docs]class LogFormatter(logging.Formatter): """ Custom formatter that adds useful information about the request to the logs when an exception happens. Cherrypy access logs are passed through without change. """
[docs] def formatException(self, exc): info = '\n'.join(( ' Request URL: %s %s' % (cherrypy.request.method.upper(), cherrypy.url()), ' Query string: ' + cherrypy.request.query_string, ' Remote IP: ' + cherrypy.request.remote.ip, ' Request UID: ' + getattr(cherrypy.request, 'girderRequestUid', '[none]') )) return ('%s\n' 'Additional info:\n' '%s' % (super().formatException(exc), info))
[docs] def format(self, record, *args, **kwargs): if hasattr(record, 'name') and hasattr(record, 'message'): if ('cherrypy.access') or'cherrypy.error')): return record.message return super().format(record, *args, **kwargs)
[docs]class StreamToLogger: """ Redirect a file-like stream to a logger. """ def __init__(self, stream, logger, level): = stream self.logger = logger self.level = level self.logger._girderLogHandlerOutput = False # This class is intended to override a default stream like sys.stdout # and sys.stderr and send that information to both the original stream # and the logger method. However, we want to preserve as much # functionality for stdout and stderr as possible, so that other # modules that send data to them can do so without a problem. The only # method we really need to override is write, but we cannot mutate the # write method on the stream itself, so we replace the stream with this # custom class. To preserve the stream methods, all of them get added # to our class instance except private and built-in methods, which, in # python, begin with _. # Fundamentally, this lets our stream replacement handle functions # flush, writeline, and others without having to enumerate them # individually. for key in dir(stream): # It's possible for a file-like object to have name appear in dir(stream) but not # actually be an attribute, thus using a default with getattr is required. # See for more. if (key != 'write' and not key.startswith('_') and (callable(getattr(stream, key, None)) or isinstance( getattr(stream, key, None), (bytes, str, int, bool)))): setattr(self, key, getattr(stream, key)) def write(self, buf): if not self.logger._girderLogHandlerOutput: self.logger._girderLogHandlerOutput = True for line in buf.rstrip().splitlines(): self.logger.log(self.level, line.rstrip()) self.logger._girderLogHandlerOutput = False
[docs]def getLogPaths(): """ Return the paths to the error and info log files. These are returned as a dict with "error" and "info" keys that point to the respective file, as well as a "root" key pointing to the log root directory. """ cfg = config.getConfig() logCfg = cfg.get('logging', {}) root = os.path.expanduser(logCfg.get('log_root', LOG_ROOT)) return { 'root': root, 'error': logCfg.get('error_log_file', os.path.join(root, 'error.log')), 'info': logCfg.get('info_log_file', os.path.join(root, 'info.log')) }
def _attachFileLogHandlers(): """ Sets up the Girder logger. """ global _quiet cfg = config.getConfig() logCfg = cfg.get('logging', {}) # If we are asked to be quiet, set a global flag so that logprint doesn't # have to get the configuration settings every time it is used. if logCfg.get('log_quiet') is True: _quiet = True logPaths = getLogPaths() # Ensure log paths are valid logDirs = [ logPaths['root'], os.path.dirname(logPaths['info']), os.path.dirname(logPaths['error']) ] for logDir in logDirs: mkdir(logDir) # Allow minimum log level to be set via config file level = logging.INFO if logCfg.get('log_level') and isinstance(getattr(logging, logCfg['log_level'], None), int): level = getattr(logging, logCfg['log_level']) logSize = MAX_LOG_SIZE if logCfg.get('log_max_size'): sizeValue = logCfg['log_max_size'] sizeUnits = {'kb': 1024, 'Mb': 1024 ** 2, 'Gb': 1024 ** 3} if sizeValue[-2:] in sizeUnits: logSize = int(sizeValue[:-2].strip()) * sizeUnits[sizeValue[-2:]] else: logSize = int(sizeValue) backupCount = int(logCfg.get('log_backup_count', LOG_BACKUP_COUNT)) fmt = LogFormatter('[%(asctime)s] %(levelname)s: %(message)s') infoMaxLevel = logging.INFO # Create log handlers if logPaths['error'] != logPaths['info']: eh = logging.handlers.RotatingFileHandler( logPaths['error'], maxBytes=logSize, backupCount=backupCount) eh.setLevel(level) eh.addFilter(LogLevelFilter(min=logging.WARNING, max=logging.CRITICAL)) eh._girderLogHandler = 'error' eh.setFormatter(fmt) logger.addHandler(eh) # Record cherrypy errors in our logs, too cherrypy.log.error_log.addHandler(eh) else: infoMaxLevel = logging.CRITICAL if isinstance(getattr(logging, logCfg.get('log_max_info_level', ''), None), int): infoMaxLevel = getattr(logging, logCfg['log_max_info_level']) ih = logging.handlers.RotatingFileHandler( logPaths['info'], maxBytes=logSize, backupCount=backupCount) ih.setLevel(level) ih.addFilter(LogLevelFilter(min=logging.DEBUG, max=infoMaxLevel)) ih._girderLogHandler = 'info' ih.setFormatter(fmt) logger.addHandler(ih) # Record cherrypy errors in our logs, too cherrypy.log.error_log.addHandler(ih) # Log http accesses to the screen and/or the info log. accessLog = logCfg.get('log_access', 'screen') if not isinstance(accessLog, (tuple, list, set)): accessLog = [accessLog] if _quiet or ('screen' not in accessLog and 'stdout' not in accessLog): cherrypy.config.update({'log.screen': False}) if 'info' in accessLog: cherrypy.log.access_log.addHandler(ih) return logger def logStdoutStderr(force=False): if _originalStdOut == sys.stdout or force: sys.stdout = StreamToLogger(_originalStdOut, logger, logging.INFO) sys.stderr = StreamToLogger(_originalStdErr, logger, logging.ERROR)
[docs]def logprint(*args, **kwargs): """ Send a message to both stdout and the appropriate logs. This behaves like Python3's print statement, plus takes additional named parameters: :param level: the log level. This determines which log handlers will store the log message. The log is always sent to stdout. :param color: one of the constants.TerminalColor values or None. :param exc_info: None to not print exception information. True for the last exception, or a tuple of exception information. """ data = io.StringIO() kwargs = (kwargs or {}).copy() level = kwargs.pop('level', logging.DEBUG) color = kwargs.pop('color', None) exc_info = kwargs.pop('exc_info', None) kwargs['file'] = data print(*args, **kwargs) data = data.getvalue().rstrip() if exc_info and not isinstance(exc_info, tuple): exc_info = sys.exc_info() data += '\n' + ''.join(traceback.format_exception(*exc_info)).rstrip() logger.log(level, data) if not _quiet: if color: data = getattr(TerminalColor, color)(data) _originalStdOut.write('%s\n' % data) _originalStdOut.flush()
def _setupCache(): """ Setup caching based on configuration file. Cache backends are forcibly replaced because Girder initially configures the regions with the null backends. """ curConfig = config.getConfig() if curConfig['cache']['enabled']: # Replace existing backend, this is necessary # because they're initially configured with the null backend cacheConfig = { '': True, 'cache.request.replace_existing_backend': True } curConfig['cache'].update(cacheConfig) cache.configure_from_config(curConfig['cache'], '') requestCache.configure_from_config(curConfig['cache'], 'cache.request.') else: # Reset caches back to null cache (in the case of server teardown) cache.configure(backend='dogpile.cache.null', replace_existing_backend=True) requestCache.configure(backend='dogpile.cache.null', replace_existing_backend=True) # Although the rateLimitBuffer has no pre-existing backend, this method may be called multiple # times in testing (where caches were already configured) rateLimitBuffer.configure(backend='dogpile.cache.memory', replace_existing_backend=True) # Expose common logging levels and colors as methods of logprint. = functools.partial(logprint, level=logging.INFO, color='info') logprint.warning = functools.partial( logprint, level=logging.WARNING, color='warning') logprint.error = functools.partial( logprint, level=logging.ERROR, color='error') logprint.success = functools.partial( logprint, level=logging.INFO, color='success') logprint.critical = functools.partial( logprint, level=logging.CRITICAL, color='error') logprint.debug = logprint logprint.exception = functools.partial( logprint, level=logging.ERROR, color='error', exc_info=True)