I have a netcdf file as in the below. I made a test to read it with the multiprocessing (so that the reading may be faster).
from netCDF4 import Dataset
import multiprocessing
fname = 'testfile.nc'
nc = Dataset(fname, 'w', format='NETCDF4')
data1 = np.random.randn(100, 100, 100)
data2 = np.random.randn(100, 100, 100)
nc.createDimension('x', 100)
nc.createDimension('y', 100)
var1 = nc.createVariable('grid1', np.float, ('x', 'y', 'z'))
var2 = nc.createVariable('grid2', np.float, ('x', 'y', 'z'))
var1[:] = data1
var2[:] = data2
nc.close()
def readnc(fname):
dataset = Dataset(fname, 'r')
return dataset['grid1'][:]
pool = multiprocessing.Pool(processes=2)
a=pool.map(readnc,fname)
pool.close()
But there is an IOERROR:
IOError: [Errno 2] No such file or directory: 'm'
The versions of netcdf and netcdf4-python are: netCDF 4.6.1, netcdf4-Python 1.4.1. I do not understand very well this problem. If someone could explain to me, it would be great ! Thanks in advance !
Best regards,
Xiaoni
I made some correction in your code, so it should work now. Nevertheless, I am not sure that you can speed up the reading of the netCDF with a code like this...
#!/usr/bin/env ipython
from netCDF4 import Dataset
import multiprocessing
import numpy as np
import time
fname = 'testfile.nc'
nc = Dataset(fname, 'w', format='NETCDF4')
data1 = np.random.randn(100, 100, 100)
data2 = np.random.randn(100, 100, 100)
nc.createDimension('x', 100)
nc.createDimension('y', 100)
nc.createDimension('z', 100)
var1 = nc.createVariable('grid1', np.float, ('x', 'y', 'z'))
var2 = nc.createVariable('grid2', np.float, ('x', 'y', 'z'))
var1[:] = data1
var2[:] = data2
nc.close()
# -------------------------------------
def readnc(fname):
#print fname
dataset = Dataset(fname, 'r')
return dataset['grid1'][:]
# -------------------------------------
for itest in range(10):
pool = multiprocessing.Pool(processes=4)
tic=time.time();
a=pool.map(readnc,(fname,))
print 'Multiprocessing: ',time.time()-tic,'seconds'
pool.close();
# --------------------------------------
for itest in range(10):
tic=time.time();
a=readnc(fname)
print 'Serial: ',time.time()-tic,'seconds'
The part with multiprocessing is taking ~0.04 seconds to run, whereas the serial part takes less than 0.01 seconds. I have a feeling that the multiprocessing part makes 4 processes with the same task - reading in the whole dataset instead of reading in only parts of it and giving one whole output with the data.
Related
I'm using a MacBook Air/OS Monterey 12.5 (There are updates available; Ventura 13.1
Python version 3.10.8 and also tried using 3.11
Pylance has pointed that all the imports I was trying to execute were not being resolved so I changed the VS Code interpreter to Python 3.10.
Anyways, here's the code:
import pandas as pd
import torch
import numpy as np
from tqdm import tqdm
from transformers import BertTokenizerFast
from transformers import BertForTokenClassification
from torch.utils.data import Dataset, DataLoader
df = pd.read_csv('ner.csv')
labels = [i.split() for i in df['labels'].values.tolist()]
unique_labels = set()
for lb in labels:
[unique_labels.add(i) for i in lb if i not in unique_labels]
# print(unique_labels)
labels_to_ids = {k: v for v, k in enumerate(sorted(unique_labels))}
ids_to_labels = {v: k for v, k in enumerate(sorted(unique_labels))}
# print(labels_to_ids)
text = df['text'].values.tolist()
example = text[36]
#print(example)
tokenizer = BertTokenizerFast.from_pretrained('bert-base-uncased')
text_tokenized = tokenizer(example, padding='max_length', max_length=512, truncation=True, return_tensors='pt')
'''
print(text_tokenized)
print(tokenizer.decode(text_tokenized.input_ids[0]))
'''
def align_label_example(tokenized_input, labels):
word_ids = tokenized_input.word_ids()
previous_word_idx = None
label_ids = []
for word_idx in word_ids:
if word_idx is None:
label_ids.append(-100)
elif word_idx != previous_word_idx:
try:
label_ids.append(labels_to_ids[labels[word_idx]])
except:
label_ids.append(-100)
else:
label_ids.append(labels_to_ids[labels[word_idx]] if label_all_tokens else -100)
previous_word_idx = word_idx
return label_ids;
label = labels[36]
label_all_tokens = False
new_label = align_label_example(text_tokenized, label)
'''
print(new_label)
print(tokenizer.convert_ids_to_tokens(text_tokenized['input_ids'][0]))
'''
def align_label(texts, labels):
tokenized_inputs = tokenizer(texts, padding='max_length', max_length=512, truncation=True)
word_ids = tokenized_inputs.word_ids()
previous_word_idx = None
label_ids = []
for word_idx in word_ids:
if word_idx is None:
label_ids.append(-100)
elif word_idx != previous_word_idx:
try:
label_ids.append(labels_to_ids[labels[word_idx]])
except:
label_ids.append(-100)
else:
try:
label_ids.append(labels_to_ids[labels[word_idx]] if label_all_tokens else -100)
except:
label_ids.append(-100)
previous_word_idx = word_idx
return label_ids
class DataSequence(torch.utils.data.Dataset):
def __init__(self, df):
lb = [i.split() for i in df['labels'].values.tolist()]
txt = df['text'].values.tolist()
self.texts = [tokenizer(str(i),
padding='max_length', max_length=512, truncation=True, return_tensors='pt') for i in txt]
self.labels = [align_label(i,j) for i,j in zip(txt, lb)]
def __len__(self):
return len(self.labels)
def get_batch_labels(self, idx):
return torch.LongTensor(self.labels[idx])
def __getitem__(self, idx):
batch_data = self.get_batch_data(idx)
batch_labels = self.get_batch_labels(idx)
return batch_data, batch_labels
df = df[0:1000]
df_train, df_val, df_test = np.split(df.sample(frac=1, random_state=42),
[int(.8 * len(df)), int(.9 * len(df))])
class BertModel(torch.nn.Module):
def __init__(self):
super(BertModel, self).__init__()
self.bert = BertForTokenClassification.from_pretrained('bert-base-cased', num_labels=len(unique_labels))
def forward(self, input_id, mask, label):
output = self.bert(input_ids=input_id, attention_mask=mask, labels=label, return_dict=False)
return output
def train_loop(model, df_train, df_val):
train_dataset = DataSequence(df_train)
val_dataset = DataSequence(df_val)
train_dataloader = DataLoader(train_dataset, num_workers=4, batch_size=BATCH_SIZE, shuffle=True)
val_dataloader = DataLoader(val_dataset, num_workers=4, batch_size=BATCH_SIZE)
use_cuda = torch.cuda.is_available()
device = torch.device('cuda' if use_cuda else 'cpu')
optimizer = torch.optim.SGD(model.parameters(), lr=LEARNING_RATE)
if use_cuda:
model = model.cuda()
best_acc = 0
best_loss = 1000
for epoch_num in range(EPOCHS):
total_acc_train = 0
total_loss_train = 0
model.train()
for train_data, train_label in tqdm(train_dataloader):
train_label = train_label.to(device)
mask = train_data['attention_mask'].squeeze(1).to(device)
input_id = train_data['input_ids'].squeeze(1).to(device)
optimizer.zero_grad()
loss, logits = model(input_id, mask, train_label)
for i in range(logits.shape[0]):
logits_clean = logits[i][train_label[i] != -100]
label_clean = train_label[i][train_label[i] != -100]
predictions = logits_clean.argmax(dim=1)
acc = (predictions == label_clean).float().mean()
total_acc_train += acc
total_loss_train += loss.item()
loss.backward()
optimizer.step()
model.eval()
total_acc_val = 0
total_loss_val = 0
for val_data, val_label in val_dataloader:
val_label = val_label.to(device)
mask = val_data['attention_mask'].squeeze(1).to(device)
input_id = val_data['input_ids'].squeeze(1).to(device)
loss, logits = model(input_id, mask, val_label)
for i in range(logits.shape[0]):
logits_clean = logits[i][val_label[i] != -100]
label_clean = val_label[i][val_label[i] != -100]
predictions = logits_clean.argmax(dim=1)
acc = (predictions == label_clean).float().mean()
total_acc_val += acc
total_loss_val += loss.item()
val_accuracy = total_acc_val / len(df_val)
val_loss = total_loss_val / len(df_val)
print(
f'Epochs: {epoch_num + 1} | Loss: {total_loss_train / len(df_train): .3f} | Accuracy: {total_acc_train / len(df_train): .3f} | Val_Loss: {total_loss_val / len(df_val): .3f} | Accuracy: {total_acc_val / len(df_val): .3f}')
LEARNING_RATE = 5e-3
EPOCHS = 5
BATCH_SIZE = 2
model = BertModel()
train_loop(model, df_train, df_val)
And the debugger says:
Exception has occurred: RuntimeError (note: full exception trace is shown but execution is paused at: <module>)
An attempt has been made to start a new process before the
current process has finished its bootstrapping phase.
This probably means that you are not using fork to start your
child processes and you have forgotten to use the proper idiom
in the main module:
if __name__ == '__main__':
freeze_support()
...
The "freeze_support()" line can be omitted if the program
is not going to be frozen to produce an executable.
File "/Users/filipedonatti/Projects/pyCodes/second_try.py", line 141, in train_loop
for train_data, train_label in tqdm(train_dataloader):
File "/Users/filipedonatti/Projects/pyCodes/second_try.py", line 197, in <module>
train_loop(model, df_train, df_val)
File "<string>", line 1, in <module> (Current frame)
By the way,
Despite using Mac, I have downloaded Anaconda-Navigator, however I've been trying and executing this code on VS Code. I've downloaded numpy, torch, datasets and other libraries through Brew with the pip3 command.
I'm at a loss, I can run the code on a google collab notebook or Jupiter notebook, and I know training models and such in my humble Mac would not be advised, but I am just exercising this so I can train and use the model in a much more powerful machine.
Please help me with this issue, I've been trying to find a solution for days.
Peace and happy holidays.
I've tried solving the issue by writing:
if __name__ == '__main__':
freeze_support()
I've tried using this:
import parallelTestModule
extractor = parallelTestModule.ParallelExtractor()
extractor.runInParallel(numProcesses=2, numThreads=4)
So...
It turns out the correct way to solve this is to implement a function to train the loop as such:
def run():
model = BertModel()
torch.multiprocessing.freeze_support()
print('loop')
train_loop(model, df_train, df_val)
if __name__ == '__main__':
run()
Redefining that train_loop line in the end. Issue solved. For more see this link: https://github.com/pytorch/pytorch/issues/5858
I'm trying to learn multiproccessing in python and I'd like to see how long my program takes to run the code by using multiproccessing, but I can't understand why the time.perf_counter function prints two really small numbers (6.00004568696022e-07 and 1.200009137392044e-06) and after that the actual amount of seconds (18.546351400000276) of the duration of the program. Can you expalin me why?
Thanks
import time
from multiprocessing import Process
start = time.perf_counter()
def counter(n):
count = 0
while count < n:
count+=1
if __name__ == '__main__':
a = Process(target = counter, args = (500000000,))
b = Process(target = counter, args = (500000000,))
a.start()
b.start()
a.join()
b.join()
print(time.perf_counter() - start)
Have the python code with pytesseract & multiprocessing. When I start the code manually from PyCharm it works fine with any number of threads. When I start the code with Win Task Sheduler with 'threads=1' it works fine.
However if I start the code with Win Task Sheduler with 'threads=2' or more than 2, it finishes without processing the images and without any errors.
I've got log messages like this. Script starts but does nothing and there is no any errors in Win logs
2020-05-24 13:09:31,834;START
2020-05-24 13:09:31,834;threads: 2
2020-05-24 13:10:31,832;START
2020-05-24 13:10:31,832;threads: 2
2020-05-24 13:11:31,851;START
2020-05-24 13:11:31,851;threads: 2
Code
from PIL import Image
import pytesseract
from pytesseract import Output
import datetime
from glob import glob
import os
import multiprocessing as multiprocessing
import cv2
import logging
def loggerinit(name, filename, overwrite):
logger = logging.getLogger(name)
logger.setLevel(logging.INFO)
# create the logging file handler
fh = logging.FileHandler(filename, encoding = 'UTF-8')
formatter = logging.Formatter('%(asctime)s;%(message)s')
fh.setFormatter(formatter)
# add handler to logger object
logger.addHandler(fh)
return logger
def getfiles(dirname, mask):
return glob(os.path.join(dirname, mask))
def tess_file(fname):
img = cv2.imread(fname)
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
im_for_T = Image.fromarray(img)
pytesseract.pytesseract.tesseract_cmd = 'C://Tesseract-OCR//tesseract.exe'
TESSDATA_PREFIX = 'C:/Program Files/Tesseract-OCR/tessdata'
try:
os.environ['OMP_THREAD_LIMIT'] = '1'
tess_data = pytesseract.image_to_osd(im_for_T, output_type=Output.DICT)
return fname, tess_data
except:
return fname, None
if __name__ == '__main__':
logger = loggerinit('tess', 'tess.log', False)
files = getfiles('Croped', '*.jpg')
t1 = datetime.datetime.now()
logger.info('START')
threads = 2
logger.info('threads: ' + str(threads))
p = multiprocessing.Pool(threads)
results = p.map(tess_file,files)
e = []
for r in results:
if type(r) == type(None):
e.append('OCR error: ' + r)
else:
print(r[0],". rotate: ",r[1]['rotate'])
p.close()
p.join()
t2 = datetime.datetime.now()
delta = (t2 - t1).total_seconds()
print('Total time: ', delta)
print('Files: ', len(files))
logger.info('Files: ' + str(len(files)))
logger.info('Stop.' + 'Total time: ' + str(delta))
# Print error if exist
for ee in e:
print(ee)
Whats wrong? How can I fix this issue?
I'm trying to parallelize my python script with the multiprocessing library. My function is part of a class and I used Pool.map.
import numpy as np
import pandas as pd
import netCDF4
import itertools
import multiprocessing as mpp
from tqdm import tqdm
Class catch2grid(object):
def __init__(self):
"""Init of catch2grid."""
self.pbar = None
...
def main(self, db_Qobs_meta_dir, ens_mean_dir, ens_sd_dir, db_Qobs_dir,
range_start, range_end):
"""Sequential computing of several flow percentiles for Qobs and Qsim,
the standard deviation of the flow percentiles of Qsim and the
KGE alpha.
db_Qobs_meta_dir -- Path to file with meta informations on the
Catchments
ens_mean_dir -- Path to file with runoff ensemble mean
ens_sd_dir -- Path to file with runoff ensemble standard deviation
db_Qobs_dir -- Path to folder with observed runoff database_Qobs_new
range_start -- starting value of range
range_end -- stopping value of range
"""
range_catch = range(range_start, range_end)
df_meta = self.import_meta(db_Qobs_meta_dir)
df_meta = self.select_catchments(df_meta)
Ens_mean, Ens_mean_Q = self.import_ens_mean(ens_mean_dir)
Ens_sd, Ens_sd_Q = self.import_ens_sd(ens_sd_dir)
Grid_lats_cen, Grid_lons_cen = self.grid_cen_arr(Ens_mean)
df_Qobs_percs = pd.DataFrame(index=range_catch, columns=
['Catch_name', 't_scale_Qobs', 'Time_cov',
'Q_5', 'Q_25', 'Q_50',
'Q_75', 'Q_95'])
df_Qsim_percs = pd.DataFrame(index=range_catch, columns=
['Catch_name', 'Q_5', 'Q_25', 'Q_50',
'Q_75', 'Q_95'])
df_sdQsim_percs = pd.DataFrame(index=range_catch, columns=
['Catch_name', 'sdQsim_5', 'sdQsim_25',
'sdQsim_50', 'sdQsim_75', 'sdQsim_95'])
df_KGE_alpha = pd.DataFrame(index=range_catch, columns=['KGE_alpha'])
df_Qobs_percs['Catch_name'] = df_meta['Catchments']\
[range_catch[0]:range_catch[-1]+1]
df_Qsim_percs['Catch_name'] = df_meta['Catchments']\
[range_catch[0]:range_catch[-1]+1]
df_sdQsim_percs['Catch_name'] = df_meta['Catchments']\
[range_catch[0]:range_catch[-1]+1]
df_KGE_alpha['Catch_name'] = df_meta['Catchments']\
[range_catch[0]:range_catch[-1]+1]
for k in range_catch:
sum_Lat_bool, sum_Lon_bool, Lat_idx, Lon_idx = self.matchgrid(df_meta,
db_Qobs_dir,
Grid_lats_cen,
Grid_lons_cen,
k)
df_Q, t_scale_Qobs = self.Qsim_to_catch(df_meta, db_Qobs_dir,
Ens_mean, Ens_mean_Q,
sum_Lat_bool, sum_Lon_bool,
Lat_idx, Lon_idx, k)
df_sdQsim = self.sdQsim_to_catch(df_meta, db_Qobs_dir, Ens_sd,
Ens_sd_Q, sum_Lat_bool,
sum_Lon_bool, Lat_idx, Lon_idx, k)
df_Qobs_percs['t_scale_Qobs'][k] = t_scale_Qobs
no_NAs = df_Q['Qobs'].isnull().sum().sum()
df_Qobs_percs['Time_cov'][k] = 1 - (no_NAs/len(df_Q.index))
df_Qobs_percs['Q_95'][k] = self.flow_perc(df_Q['Qobs'], perc=95)
df_Qobs_percs['Q_75'][k] = self.flow_perc(df_Q['Qobs'], perc=75)
df_Qobs_percs['Q_50'][k] = self.flow_perc(df_Q['Qobs'], perc=50)
df_Qobs_percs['Q_25'][k] = self.flow_perc(df_Q['Qobs'], perc=25)
df_Qobs_percs['Q_5'][k] = self.flow_perc(df_Q['Qobs'], perc=5)
df_Qsim_percs['Q_95'][k] = self.flow_perc(df_Q['Qsim'], perc=95)
df_Qsim_percs['Q_75'][k] = self.flow_perc(df_Q['Qsim'], perc=75)
df_Qsim_percs['Q_50'][k] = self.flow_perc(df_Q['Qsim'], perc=50)
df_Qsim_percs['Q_25'][k] = self.flow_perc(df_Q['Qsim'], perc=25)
df_Qsim_percs['Q_5'][k] = self.flow_perc(df_Q['Qsim'], perc=5)
df_sdQsim_percs['sdQsim_95'][k] = self.flow_perc_sd(df_Q['Qsim'], df_sdQsim['sdQsim'], perc=95)
df_sdQsim_percs['sdQsim_75'][k] = self.flow_perc_sd(df_Q['Qsim'], df_sdQsim['sdQsim'], perc=75)
df_sdQsim_percs['sdQsim_50'][k] = self.flow_perc_sd(df_Q['Qsim'], df_sdQsim['sdQsim'], perc=50)
df_sdQsim_percs['sdQsim_25'][k] = self.flow_perc_sd(df_Q['Qsim'], df_sdQsim['sdQsim'], perc=25)
df_sdQsim_percs['sdQsim_5'][k] = self.flow_perc_sd(df_Q['Qsim'], df_sdQsim['sdQsim'], perc=5)
df_KGE_alpha['KGE_alpha'][k] = self.KGE_alpha(df_Q['Qsim'], df_Q['Qobs'])
# display progress
self.pbar.update(1)
df_Qobs_percs.index = df_Qobs_percs['Catch_name']
df_Qsim_percs.index = df_Qsim_percs['Catch_name']
df_sdQsim_percs.index = df_sdQsim_percs['Catch_name']
df_KGE_alpha.index = df_KGE_alpha['Catch_name']
df_Qobs_percs = df_Qobs_percs.loc[:, 'Q_5':'Q_95']
df_Qsim_percs = df_Qsim_percs.loc[:, 'Q_5':'Q_95']
df_sdQsim_percs = df_sdQsim_percs.loc[:, 'sdQsim_5':'sdQsim_95']
df_KGE_alpha = df_KGE_alpha.loc[:, 'KGE_alpha']
return df_Qobs_percs, df_Qsim_percs, df_sdQsim_percs, df_KGE_alpha
def main_par(self, db_Qobs_meta_dir, ens_mean_dir, ens_sd_dir, db_Qobs_dir):
"""Parallel computing of several flow percentiles for Qobs and Qsim,
the standard deviation of the flow percentiles of Qsim and the
KGE alpha.
db_Qobs_meta_dir -- Path to file with meta informations on the
Catchments
ens_mean_dir -- Path to file with runoff ensemble mean
ens_sd_dir -- Path to file with runoff ensemble standard deviation
db_Qobs_dir -- Path to folder with observed runoff database_Qobs_new
"""
cpu_cores = mpp.cpu_count() - 1
df_meta = self.import_meta(db_Qobs_meta_dir)
df_meta = self.select_catchments(df_meta)
# chunking subsets for parallelization
ll_start = []
ll_end = []
lin_dist = np.linspace(0, len(df_meta.index), cpu_cores+1)
l = len(lin_dist)
# list of tuples with input arguments for map
for i in range(len(lin_dist) - 1):
temp = list(range(int(lin_dist[i]), int(lin_dist[i+1]), 1))
ll_start.append(temp[0])
ll_end.append(temp[-1]+1)
ll_db_Qobs_meta_dir = list(itertools.repeat(db_Qobs_meta_dir, l))
ll_Ens_mean_dir = list(itertools.repeat(ens_mean_dir, l))
ll_Ens_sd_dir = list(itertools.repeat(ens_sd_dir, l))
ll_db_Qobs_dir = list(itertools.repeat(db_Qobs_dir, l))
subsets = zip(ll_db_Qobs_meta_dir, ll_Ens_mean_dir, ll_Ens_sd_dir,
ll_db_Qobs_dir, ll_start, ll_end)
p = mpp.Pool(cpu_cores) # launch pool of workers
res = p.starmap(self.main, subsets)
p.close()
p.join()
res_obs = []
res_sim = []
res_simsd = []
res_kgealpha = []
# collect dataframes and merge them
[res_obs.append(res[:][i][0]) for i in range(len(res))]
[res_sim.append(res[:][i][1]) for i in range(len(res))]
[res_simsd.append(res[:][i][2]) for i in range(len(res))]
[res_kgealpha.append(res[:][i][3]) for i in range(len(res))]
df_Qobs_percs = pd.concat(res_obs[:], ignore_index=True)
df_Qsim_percs = pd.concat(res_sim[:], ignore_index=True)
df_sdQsim_percs = pd.concat(res_simsd[:], ignore_index=True)
df_KGE_alpha = pd.concat(res_kgealpha[:], ignore_index=True)
return df_Qobs_percs, df_Qsim_percs, df_sdQsim_percs, df_KGE_alpha
...
if __name__ == "__main__":
cpu_cores = mp.cpu_count() - 1
c2g = catch2grid()
p = mp.Pool(cpu_cores) # launch pool of workers
c2g.init_pbar(l)
ll_range_catch = list(range(0, 5000))
res = p.map(c2g.main_par, ll_range_catch)
p.close()
p.join()
After running it the following error message is displayed:
File "<ipython-input-1-3828921ab3bd>", line 1, in <module>
runfile('/Users/robinschwemmle/Desktop/MSc_Thesis/Python/catch2grid.py', wdir='/Users/robinschwemmle/Desktop/MSc_Thesis/Python')
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/site-packages/spyder/utils/site/sitecustomize.py", line 705, in runfile
execfile(filename, namespace)
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/site-packages/spyder/utils/site/sitecustomize.py", line 102, in execfile
exec(compile(f.read(), filename, 'exec'), namespace)
File "/Users/robinschwemmle/Desktop/MSc_Thesis/Python/catch2grid.py", line 1285, in <module>
c2g.main_par(db_Qobs_meta_dir, Ens_mean_dir, Ens_sd_dir, db_Qobs_dir)
File "/Users/robinschwemmle/Desktop/MSc_Thesis/Python/catch2grid.py", line 798, in main_par
res = p.starmap(self.main, subsets)
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/multiprocessing/pool.py", line 274, in starmap
return self._map_async(func, iterable, starmapstar, chunksize).get()
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/multiprocessing/pool.py", line 644, in get
raise self._value
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/multiprocessing/pool.py", line 424, in _handle_tasks
put(task)
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/multiprocessing/connection.py", line 206, in send
self._send_bytes(_ForkingPickler.dumps(obj))
File "/Users/robinschwemmle/anaconda/envs/py36/lib/python3.6/multiprocessing/reduction.py", line 51, in dumps
cls(buf, protocol).dump(obj)
PicklingError: Can't pickle <function <lambda> at 0x1164e42f0>: attribute lookup <lambda> on jupyter_client.session failed
The error occured just a few days ago. Before the code was working properly. Have there been any changes to the mulitprocessing or pickling library I'm not aware of? Or has anyone an advice for me which parallel library I could choose instead?
I'm working with Graphite monitoring using Carbon and Ceres as the storage method. I have some problems with correcting bad data. It seems that (due to various problems) I've ended up with overlapping files. That is, since Carbon / Ceres stores the data as timestamp#interval.slice, I can have two or more files with overlapping time ranges.
There are two kinds of overlaps:
File A: +------------+ orig file
File B: +-----+ subset
File C: +---------+ overlap
This is causing problems because the existing tools available (ceres-maintenance defrag and rollup) don't cope with these overlaps. Instead, they skip the directory and move on. This is a problem, obviously.
I've created a script that fixes this problem, as follows:
For subsets, just delete the subset file.
For overlaps, using the file system 'truncate' on the orig file at the point where the next file starts. While it is possible to cut off the start of the overlap file and rename it properly, I would suggest that this is fraught with danger.
I've found that it's possible to do this in two ways:
Walk the dirs and iterate over the files, fixing as you go, and find the file subsets, remove them;
Walk the dirs and fix all the problems in a dir before moving on. This is BY FAR the faster approach, since the dir walk is hugely time consuming.
Code:
#!/usr/bin/env python2.6
################################################################################
import io
import os
import time
import sys
import string
import logging
import unittest
import datetime
import random
import zmq
import json
import socket
import traceback
import signal
import select
import simplejson
import cPickle as pickle
import re
import shutil
import collections
from pymongo import Connection
from optparse import OptionParser
from pprint import pprint, pformat
################################################################################
class SliceFile(object):
def __init__(self, fname):
self.name = fname
basename = fname.split('/')[-1]
fnArray = basename.split('#')
self.timeStart = int(fnArray[0])
self.freq = int(fnArray[1].split('.')[0])
self.size = None
self.numPoints = None
self.timeEnd = None
self.deleted = False
def __repr__(self):
out = "Name: %s, tstart=%s tEnd=%s, freq=%s, size=%s, npoints=%s." % (
self.name, self.timeStart, self.timeEnd, self.freq, self.size, self.numPoints)
return out
def setVars(self):
self.size = os.path.getsize(self.name)
self.numPoints = int(self.size / 8)
self.timeEnd = self.timeStart + (self.numPoints * self.freq)
################################################################################
class CeresOverlapFixup(object):
def __del__(self):
import datetime
self.writeLog("Ending at %s" % (str(datetime.datetime.today())))
self.LOGFILE.flush()
self.LOGFILE.close()
def __init__(self):
self.verbose = False
self.debug = False
self.LOGFILE = open("ceresOverlapFixup.log", "a")
self.badFilesList = set()
self.truncated = 0
self.subsets = 0
self.dirsExamined = 0
self.lastStatusTime = 0
def getOptionParser(self):
return OptionParser()
def getOptions(self):
parser = self.getOptionParser()
parser.add_option("-d", "--debug", action="store_true", dest="debug", default=False, help="debug mode for this program, writes debug messages to logfile." )
parser.add_option("-v", "--verbose", action="store_true", dest="verbose", default=False, help="verbose mode for this program, prints a lot to stdout." )
parser.add_option("-b", "--basedir", action="store", type="string", dest="basedir", default=None, help="base directory location to start converting." )
(options, args) = parser.parse_args()
self.debug = options.debug
self.verbose = options.verbose
self.basedir = options.basedir
assert self.basedir, "must provide base directory."
# Examples:
# ./updateOperations/1346805360#60.slice
# ./updateOperations/1349556660#60.slice
# ./updateOperations/1346798040#60.slice
def getFileData(self, inFilename):
ret = SliceFile(inFilename)
ret.setVars()
return ret
def removeFile(self, inFilename):
os.remove(inFilename)
#self.writeLog("removing file: %s" % (inFilename))
self.subsets += 1
def truncateFile(self, fname, newSize):
if self.verbose:
self.writeLog("Truncating file, name=%s, newsize=%s" % (pformat(fname), pformat(newSize)))
IFD = None
try:
IFD = os.open(fname, os.O_RDWR|os.O_CREAT)
os.ftruncate(IFD, newSize)
os.close(IFD)
self.truncated += 1
except:
self.writeLog("Exception during truncate: %s" % (traceback.format_exc()))
try:
os.close(IFD)
except:
pass
return
def printStatus(self):
now = self.getNowTime()
if ((now - self.lastStatusTime) > 10):
self.writeLog("Status: time=%d, Walked %s dirs, subsetFilesRemoved=%s, truncated %s files." % (now, self.dirsExamined, self.subsets, self.truncated))
self.lastStatusTime = now
def fixupThisDir(self, inPath, inFiles):
# self.writeLog("Fixing files in dir: %s" % (inPath))
if not '.ceres-node' in inFiles:
# self.writeLog("--> Not a slice directory, skipping.")
return
self.dirsExamined += 1
sortedFiles = sorted(inFiles)
sortedFiles = [x for x in sortedFiles if ((x != '.ceres-node') and (x.count('#') > 0)) ]
lastFile = None
fileObjList = []
for thisFile in sortedFiles:
wholeFilename = os.path.join(inPath, thisFile)
try:
curFile = self.getFileData(wholeFilename)
fileObjList.append(curFile)
except:
self.badFilesList.add(wholeFilename)
self.writeLog("ERROR: file %s, %s" % (wholeFilename, traceback.format_exc()))
# name is timeStart, really.
fileObjList = sorted(fileObjList, key=lambda thisObj: thisObj.name)
while fileObjList:
self.printStatus()
changes = False
firstFile = fileObjList[0]
removedFiles = []
for curFile in fileObjList[1:]:
if (curFile.timeEnd <= firstFile.timeEnd):
# have subset file. elim.
self.removeFile(curFile.name)
removedFiles.append(curFile.name)
self.subsets += 1
changes = True
if self.verbose:
self.writeLog("Subset file situation. First=%s, overlap=%s" % (firstFile, curFile))
fileObjList = [x for x in fileObjList if x.name not in removedFiles]
if (len(fileObjList) < 2):
break
secondFile = fileObjList[1]
# LT is right. FirstFile's timeEnd is always the first open time after first is done.
# so, first starts#100, len=2, end=102, positions used=100,101. second start#102 == OK.
if (secondFile.timeStart < firstFile.timeEnd):
# truncate first file.
# file_A (last): +---------+
# file_B (curr): +----------+
# solve by truncating previous file at startpoint of current file.
newLenFile_A_seconds = int(secondFile.timeStart - firstFile.timeStart)
newFile_A_datapoints = int(newLenFile_A_seconds / firstFile.freq)
newFile_A_bytes = int(newFile_A_datapoints) * 8
if (not newFile_A_bytes):
fileObjList = fileObjList[1:]
continue
assert newFile_A_bytes, "Must have size. newLenFile_A_seconds=%s, newFile_A_datapoints=%s, newFile_A_bytes=%s." % (newLenFile_A_seconds, newFile_A_datapoints, newFile_A_bytes)
self.truncateFile(firstFile.name, newFile_A_bytes)
if self.verbose:
self.writeLog("Truncate situation. First=%s, overlap=%s" % (firstFile, secondFile))
self.truncated += 1
fileObjList = fileObjList[1:]
changes = True
if not changes:
fileObjList = fileObjList[1:]
def getNowTime(self):
return time.time()
def walkDirStructure(self):
startTime = self.getNowTime()
self.lastStatusTime = startTime
updateStatsDict = {}
self.okayFiles = 0
emptyFiles = 0
for (thisPath, theseDirs, theseFiles) in os.walk(self.basedir):
self.printStatus()
self.fixupThisDir(thisPath, theseFiles)
self.dirsExamined += 1
endTime = time.time()
# time.sleep(11)
self.printStatus()
self.writeLog( "now = %s, started at %s, elapsed time = %s seconds." % (startTime, endTime, endTime - startTime))
self.writeLog( "Done.")
def writeLog(self, instring):
print instring
print >> self.LOGFILE, instring
self.LOGFILE.flush()
def main(self):
self.getOptions()
self.walkDirStructure()