X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/4dac499d30988fd8289a40ad4128d6d471037180..81693ee04f18e68558c9206705ef589cca2460b0:/sdk/cwl/arvados_cwl/fsaccess.py diff --git a/sdk/cwl/arvados_cwl/fsaccess.py b/sdk/cwl/arvados_cwl/fsaccess.py index 7736cab4d8..3744b4a93a 100644 --- a/sdk/cwl/arvados_cwl/fsaccess.py +++ b/sdk/cwl/arvados_cwl/fsaccess.py @@ -1,10 +1,21 @@ +# Copyright (C) The Arvados Authors. All rights reserved. +# +# SPDX-License-Identifier: Apache-2.0 + +from future import standard_library +standard_library.install_aliases() +from builtins import object +from builtins import str +from future.utils import viewvalues + import fnmatch import os import errno -import urlparse +import urllib.parse import re import logging import threading +from collections import OrderedDict import ruamel.yaml as yaml @@ -17,24 +28,60 @@ import arvados.collection import arvados.arvfile import arvados.errors +from googleapiclient.errors import HttpError + from schema_salad.ref_resolver import DefaultFetcher logger = logging.getLogger('arvados.cwl-runner') +pdh_size = re.compile(r'([0-9a-f]{32})\+(\d+)(\+\S+)*') + class CollectionCache(object): - def __init__(self, api_client, keep_client, num_retries): + def __init__(self, api_client, keep_client, num_retries, + cap=256*1024*1024, + min_entries=2): self.api_client = api_client self.keep_client = keep_client - self.collections = {} + self.num_retries = num_retries + self.collections = OrderedDict() self.lock = threading.Lock() + self.total = 0 + self.cap = cap + self.min_entries = min_entries + + def set_cap(self, cap): + self.cap = cap + + def cap_cache(self, required): + # ordered dict iterates from oldest to newest + for pdh, v in list(self.collections.items()): + available = self.cap - self.total + if available >= required or len(self.collections) < self.min_entries: + return + # cut it loose + logger.debug("Evicting collection reader %s from cache (cap %s total %s required %s)", pdh, self.cap, self.total, required) + del self.collections[pdh] + self.total -= v[1] def get(self, pdh): with self.lock: if pdh not in self.collections: + m = pdh_size.match(pdh) + if m: + self.cap_cache(int(m.group(2)) * 128) logger.debug("Creating collection reader for %s", pdh) - self.collections[pdh] = arvados.collection.CollectionReader(pdh, api_client=self.api_client, - keep_client=self.keep_client) - return self.collections[pdh] + cr = arvados.collection.CollectionReader(pdh, api_client=self.api_client, + keep_client=self.keep_client, + num_retries=self.num_retries) + sz = len(cr.manifest_text()) * 128 + self.collections[pdh] = (cr, sz) + self.total += sz + else: + cr, sz = self.collections[pdh] + # bump it to the back + del self.collections[pdh] + self.collections[pdh] = (cr, sz) + return cr class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): @@ -49,7 +96,7 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): p = sp[0] if p.startswith("keep:") and arvados.util.keep_locator_pattern.match(p[5:]): pdh = p[5:] - return (self.collection_cache.get(pdh), sp[1] if len(sp) == 2 else None) + return (self.collection_cache.get(pdh), urllib.parse.unquote(sp[1]) if len(sp) == 2 else None) else: return (None, path) @@ -77,21 +124,27 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): def glob(self, pattern): collection, rest = self.get_collection(pattern) - if collection and not rest: + if collection is not None and not rest: return [pattern] patternsegments = rest.split("/") - return self._match(collection, patternsegments, "keep:" + collection.manifest_locator()) + return sorted(self._match(collection, patternsegments, "keep:" + collection.manifest_locator())) def open(self, fn, mode): collection, rest = self.get_collection(fn) - if collection: + if collection is not None: return collection.open(rest, mode) else: return super(CollectionFsAccess, self).open(self._abs(fn), mode) def exists(self, fn): - collection, rest = self.get_collection(fn) - if collection: + try: + collection, rest = self.get_collection(fn) + except HttpError as err: + if err.resp.status == 404: + return False + else: + raise + if collection is not None: if rest: return collection.exists(rest) else: @@ -99,9 +152,20 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): else: return super(CollectionFsAccess, self).exists(fn) + def size(self, fn): # type: (unicode) -> bool + collection, rest = self.get_collection(fn) + if collection is not None: + if rest: + arvfile = collection.find(rest) + if isinstance(arvfile, arvados.arvfile.ArvadosFile): + return arvfile.size() + raise IOError(errno.EINVAL, "Not a path to a file %s" % (fn)) + else: + return super(CollectionFsAccess, self).size(fn) + def isfile(self, fn): # type: (unicode) -> bool collection, rest = self.get_collection(fn) - if collection: + if collection is not None: if rest: return isinstance(collection.find(rest), arvados.arvfile.ArvadosFile) else: @@ -111,7 +175,7 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): def isdir(self, fn): # type: (unicode) -> bool collection, rest = self.get_collection(fn) - if collection: + if collection is not None: if rest: return isinstance(collection.find(rest), arvados.collection.RichCollectionBase) else: @@ -121,7 +185,7 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): def listdir(self, fn): # type: (unicode) -> List[unicode] collection, rest = self.get_collection(fn) - if collection: + if collection is not None: if rest: dir = collection.find(rest) else: @@ -130,7 +194,7 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): raise IOError(errno.ENOENT, "Directory '%s' in '%s' not found" % (rest, collection.portable_data_hash())) if not isinstance(dir, arvados.collection.RichCollectionBase): raise IOError(errno.ENOENT, "Path '%s' in '%s' is not a Directory" % (rest, collection.portable_data_hash())) - return [abspath(l, fn) for l in dir.keys()] + return [abspath(l, fn) for l in list(dir.keys())] else: return super(CollectionFsAccess, self).listdir(fn) @@ -143,22 +207,19 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess): if path.startswith("$(task.tmpdir)") or path.startswith("$(task.outdir)"): return path collection, rest = self.get_collection(path) - if collection: + if collection is not None: return path else: return os.path.realpath(path) class CollectionFetcher(DefaultFetcher): - def __init__(self, cache, session, api_client=None, fs_access=None, num_retries=4, overrides=None): + def __init__(self, cache, session, api_client=None, fs_access=None, num_retries=4): super(CollectionFetcher, self).__init__(cache, session) self.api_client = api_client self.fsaccess = fs_access self.num_retries = num_retries - self.overrides = overrides if overrides else {} def fetch_text(self, url): - if url in self.overrides: - return self.overrides[url] if url.startswith("keep:"): with self.fsaccess.open(url, "r") as f: return f.read() @@ -169,8 +230,6 @@ class CollectionFetcher(DefaultFetcher): return super(CollectionFetcher, self).fetch_text(url) def check_exists(self, url): - if url in self.overrides: - return True try: if url.startswith("http://arvados.org/cwl"): return True @@ -181,8 +240,8 @@ class CollectionFetcher(DefaultFetcher): return True except arvados.errors.NotFoundError: return False - except: - logger.exception("Got unexpected exception checking if file exists:") + except Exception: + logger.exception("Got unexpected exception checking if file exists") return False return super(CollectionFetcher, self).check_exists(url) @@ -190,11 +249,11 @@ class CollectionFetcher(DefaultFetcher): if not url: return base_url - urlsp = urlparse.urlsplit(url) + urlsp = urllib.parse.urlsplit(url) if urlsp.scheme or not base_url: return url - basesp = urlparse.urlsplit(base_url) + basesp = urllib.parse.urlsplit(base_url) if basesp.scheme in ("keep", "arvwf"): if not basesp.path: raise IOError(errno.EINVAL, "Invalid Keep locator", base_url) @@ -215,27 +274,36 @@ class CollectionFetcher(DefaultFetcher): baseparts.pop() path = "/".join([pdh] + baseparts + urlparts) - return urlparse.urlunsplit((basesp.scheme, "", path, "", urlsp.fragment)) + return urllib.parse.urlunsplit((basesp.scheme, "", path, "", urlsp.fragment)) return super(CollectionFetcher, self).urljoin(base_url, url) + schemes = [u"file", u"http", u"https", u"mailto", u"keep", u"arvwf"] + + def supported_schemes(self): # type: () -> List[Text] + return self.schemes + + workflow_uuid_pattern = re.compile(r'[a-z0-9]{5}-7fd4e-[a-z0-9]{15}') pipeline_template_uuid_pattern = re.compile(r'[a-z0-9]{5}-p5p6p-[a-z0-9]{15}') def collectionResolver(api_client, document_loader, uri, num_retries=4): + if uri.startswith("keep:") or uri.startswith("arvwf:"): + return str(uri) + if workflow_uuid_pattern.match(uri): - return "arvwf:%s#main" % (uri) + return u"arvwf:%s#main" % (uri) if pipeline_template_uuid_pattern.match(uri): pt = api_client.pipeline_templates().get(uuid=uri).execute(num_retries=num_retries) - return "keep:" + pt["components"].values()[0]["script_parameters"]["cwl:tool"] + return u"keep:" + viewvalues(pt["components"])[0]["script_parameters"]["cwl:tool"] p = uri.split("/") if arvados.util.keep_locator_pattern.match(p[0]): - return "keep:%s" % (uri) + return u"keep:%s" % (uri) if arvados.util.collection_uuid_pattern.match(p[0]): - return "keep:%s%s" % (api_client.collections(). + return u"keep:%s%s" % (api_client.collections(). get(uuid=p[0]).execute()["portable_data_hash"], uri[len(p[0]):])