X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/6e17218f2fc39a828af4d3a19bdb29243739dafd..432ec6a698a5ff2616541745aa7afa9074b970c3:/sdk/cwl/arvados_cwl/http.py diff --git a/sdk/cwl/arvados_cwl/http.py b/sdk/cwl/arvados_cwl/http.py index ab59ad3a53..f2415bcffe 100644 --- a/sdk/cwl/arvados_cwl/http.py +++ b/sdk/cwl/arvados_cwl/http.py @@ -1,3 +1,11 @@ +# Copyright (C) The Arvados Authors. All rights reserved. +# +# SPDX-License-Identifier: Apache-2.0 + +from __future__ import division +from future import standard_library +standard_library.install_aliases() + import requests import email.utils import time @@ -5,18 +13,30 @@ import datetime import re import arvados import arvados.collection -import urlparse +import urllib.parse import logging +import calendar +import urllib.parse logger = logging.getLogger('arvados.cwl-runner') def my_formatdate(dt): - return email.utils.formatdate(timeval=time.mktime(now.timetuple()), localtime=False, usegmt=True) + return email.utils.formatdate(timeval=calendar.timegm(dt.timetuple()), + localtime=False, usegmt=True) def my_parsedate(text): - return datetime.datetime(*email.utils.parsedate(text)[:6]) - -def fresh_cache(url, properties): + parsed = email.utils.parsedate_tz(text) + if parsed: + if parsed[9]: + # Adjust to UTC + return datetime.datetime(*parsed[:6]) + datetime.timedelta(seconds=parsed[9]) + else: + # TZ is zero or missing, assume UTC. + return datetime.datetime(*parsed[:6]) + else: + return datetime.datetime(1970, 1, 1) + +def fresh_cache(url, properties, now): pr = properties[url] expires = None @@ -41,74 +61,163 @@ def fresh_cache(url, properties): if not expires: return False - return (datetime.datetime.utcnow() < expires) + return (now < expires) -def remember_headers(url, properties, headers): +def remember_headers(url, properties, headers, now): properties.setdefault(url, {}) for h in ("Cache-Control", "ETag", "Expires", "Date", "Content-Length"): if h in headers: properties[url][h] = headers[h] if "Date" not in headers: - properties[url]["Date"] = my_formatdate(datetime.datetime.utcnow()) + properties[url]["Date"] = my_formatdate(now) -def changed(url, properties): - req = requests.head(url) - remember_headers(url, properties, req.headers) +def changed(url, clean_url, properties, now): + req = requests.head(url, allow_redirects=True) if req.status_code != 200: - raise Exception("Got status %s" % req.status_code) + # Sometimes endpoints are misconfigured and will deny HEAD but + # allow GET so instead of failing here, we'll try GET If-None-Match + return True + + etag = properties[url].get("ETag") + + if url in properties: + del properties[url] + remember_headers(clean_url, properties, req.headers, now) + + if "ETag" in req.headers and etag == req.headers["ETag"]: + # Didn't change + return False - pr = properties[url] - if "ETag" in pr and "ETag" in req.headers: - if pr["ETag"] == req.headers["ETag"]: - return False return True -def http_to_keep(api, project_uuid, url): - r = api.collections().list(filters=[["properties", "exists", url]]).execute() - name = urlparse.urlparse(url).path.split("/")[-1] +def etag_quote(etag): + # if it already has leading and trailing quotes, do nothing + if etag[0] == '"' and etag[-1] == '"': + return etag + else: + # Add quotes. + return '"' + etag + '"' + + +def http_to_keep(api, project_uuid, url, utcnow=datetime.datetime.utcnow, varying_url_params="", prefer_cached_downloads=False): + varying_params = [s.strip() for s in varying_url_params.split(",")] + + parsed = urllib.parse.urlparse(url) + query = [q for q in urllib.parse.parse_qsl(parsed.query) + if q[0] not in varying_params] + + clean_url = urllib.parse.urlunparse((parsed.scheme, parsed.netloc, parsed.path, parsed.params, + urllib.parse.urlencode(query, safe="/"), parsed.fragment)) - for item in r["items"]: + r1 = api.collections().list(filters=[["properties", "exists", url]]).execute() + + if clean_url == url: + items = r1["items"] + else: + r2 = api.collections().list(filters=[["properties", "exists", clean_url]]).execute() + items = r1["items"] + r2["items"] + + now = utcnow() + + etags = {} + + for item in items: properties = item["properties"] - if fresh_cache(url, properties): - # Do nothing - return "keep:%s/%s" % (item["portable_data_hash"], name) - if not changed(url, properties): + if clean_url in properties: + cache_url = clean_url + elif url in properties: + cache_url = url + else: + return False + + if prefer_cached_downloads or fresh_cache(cache_url, properties, now): + # HTTP caching rules say we should use the cache + cr = arvados.collection.CollectionReader(item["portable_data_hash"], api_client=api) + return "keep:%s/%s" % (item["portable_data_hash"], list(cr.keys())[0]) + + if not changed(cache_url, clean_url, properties, now): # ETag didn't change, same content, just update headers api.collections().update(uuid=item["uuid"], body={"collection":{"properties": properties}}).execute() - return "keep:%s/%s" % (item["portable_data_hash"], name) - - properties = {} - req = requests.get(url, stream=True) + cr = arvados.collection.CollectionReader(item["portable_data_hash"], api_client=api) + return "keep:%s/%s" % (item["portable_data_hash"], list(cr.keys())[0]) - if req.status_code != 200: - raise Exception("Failed to download '%s' got status %s " % (req.status_code, url)) + if "ETag" in properties[cache_url] and len(properties[cache_url]["ETag"]) > 2: + etags[properties[cache_url]["ETag"]] = item - remember_headers(url, properties, req.headers) + logger.debug("Found ETags %s", etags) - logger.info("Downloading %s (%s bytes)", url, properties[url]["Content-Length"]) + properties = {} + headers = {} + if etags: + headers['If-None-Match'] = ', '.join([etag_quote(k) for k,v in etags.items()]) + logger.debug("Sending GET request with headers %s", headers) + req = requests.get(url, stream=True, allow_redirects=True, headers=headers) + + if req.status_code not in (200, 304): + raise Exception("Failed to download '%s' got status %s " % (url, req.status_code)) + + remember_headers(clean_url, properties, req.headers, now) + + if req.status_code == 304 and "ETag" in req.headers and req.headers["ETag"] in etags: + item = etags[req.headers["ETag"]] + item["properties"].update(properties) + api.collections().update(uuid=item["uuid"], body={"collection":{"properties": item["properties"]}}).execute() + cr = arvados.collection.CollectionReader(item["portable_data_hash"], api_client=api) + return "keep:%s/%s" % (item["portable_data_hash"], list(cr.keys())[0]) + + if "Content-Length" in properties[clean_url]: + cl = int(properties[clean_url]["Content-Length"]) + logger.info("Downloading %s (%s bytes)", url, cl) + else: + cl = None + logger.info("Downloading %s (unknown size)", url) c = arvados.collection.Collection() + if req.headers.get("Content-Disposition"): + grp = re.search(r'filename=("((\"|[^"])+)"|([^][()<>@,;:\"/?={} ]+))', req.headers["Content-Disposition"]) + if grp.group(2): + name = grp.group(2) + else: + name = grp.group(4) + else: + name = parsed.path.split("/")[-1] + count = 0 start = time.time() checkpoint = start - with c.open(name, "w") as f: + with c.open(name, "wb") as f: for chunk in req.iter_content(chunk_size=1024): count += len(chunk) f.write(chunk) - now = time.time() - if (now - checkpoint) > 20: - bps = (float(count)/float(now - start)) - logger.info("%2.1f%% complete, %3.2f MiB/s, %1.0f seconds left", - float(count * 100) / float(properties[url]["Content-Length"]), - bps/(1024*1024), - (int(properties[url]["Content-Length"])-count)/bps) - checkpoint = now - - c.save_new(name="Downloaded from %s" % url, owner_uuid=project_uuid, ensure_unique_name=True) + loopnow = time.time() + if (loopnow - checkpoint) > 20: + bps = count / (loopnow - start) + if cl is not None: + logger.info("%2.1f%% complete, %3.2f MiB/s, %1.0f seconds left", + ((count * 100) / cl), + (bps // (1024*1024)), + ((cl-count) // bps)) + else: + logger.info("%d downloaded, %3.2f MiB/s", count, (bps / (1024*1024))) + checkpoint = loopnow + + logger.info("Download complete") + + collectionname = "Downloaded from %s" % urllib.parse.quote(clean_url, safe='') + + # max length - space to add a timestamp used by ensure_unique_name + max_name_len = 254 - 28 + + if len(collectionname) > max_name_len: + over = len(collectionname) - max_name_len + split = int(max_name_len/2) + collectionname = collectionname[0:split] + "…" + collectionname[split+over:] + + c.save_new(name=collectionname, owner_uuid=project_uuid, ensure_unique_name=True) api.collections().update(uuid=c.manifest_locator(), body={"collection":{"properties": properties}}).execute()